THESIS
ASSESSING IMPACTS OF RAINFALL PATTERNS, POPULATION GROWTH, AND SEA
LEVEL RISE ON GROUNDWATER SUPPLY IN THE REPUBLIC OF MALDIVES
Submitted by
Chenda Deng
Department of Civil and Environmental Engineering
In partial fulfillment of the requirements
For the Degree of Master of Science
Colorado State University
Fort Collins, Colorado
Summer 2016
Master’s Committee:
Advisor: Ryan Bailey
Neil Grigg
William E. Sanford
Copyright by Chenda Deng 2016
All Rights Reserved
ABSTRACT
ASSESSING IMPACTS OF RAINFALL PATTERNS, POPULATION GROWTH, AND SEA
LEVEL RISE ON GROUNDWATER SUPPLY IN THE REPUBLIC OF MALDIVES
Groundwater resources of the Republic of the Maldives are threatened by a variety of factors
including variable future rainfall patterns, continued population growth and associated pumping
demands, rising sea level, and contamination from the land surface. The Maldives is composed
of approximately 2,000 coral islands residing in 26 atolls in the Indian Ocean, with each coral
island less than a few square kilometers in surface area and less than a few meters in elevation.
This thesis uses numerical modeling techniques to assess the influence of variable rainfall
patterns, increased pumping due to population growth, and sea level rise on fresh groundwater
supply of the coral islands that comprise the Maldives. The density-dependent groundwater flow
and solute transport model SUTRA (Saturated Unsaturated Transport) is used for all simulations,
with the model simulating the spatial extent of the freshwater lens in the aquifer of the coral
islands.
The thesis first assesses changes in groundwater supply due to variable rainfall patterns in the
coming decades, a key component of water resources management for the country. Using a suite
of two-dimensional vertical cross-section models, time-dependent thickness of the freshwater
lens is simulated for a range of island sizes (200 m to 1100 m) during the time period of 2011 to
2050, with recharge to the freshwater lens calculated using rainfall patterns provided by General
Circulation Models (GCM) for the three distinct geographic regions (north, central, south) of the
Maldives. Results show that average lens thickness of islands in all three geographic regions
ii
during the 2031-2050 time period is slightly greater than during the 2011-2030 time period,
indicating a mild increase in future available groundwater supply under predicted conditions.
Average lens thickness during 2011-2030 for islands of 200 m, 400 m, 600 m, and 1100 m width
is 0.5 m, 3.0 m, 7.0 m, and 12.2 m, respectively, with these values increasing by 1-5% during
2031-2050 time period. However, these results do not include the effect of sea level rise.
To quantify the total available groundwater on a representative island and to provide accurate
simulation of the effect of radial pumping on the freshwater lens, a three dimensional model is
created for the island of Gan (Area: 598 ha, Population: 4,280) to evaluate the impact of
increasing pumping and sea-level rise on future groundwater resources. Simulations covering the
2012-2050 period are used to compare scenarios of future rainfall, pumping vs. non-pumping,
varying rates of population growth and hence of groundwater pumping, and sea level rise (0.5 m
by 2100) vs. no sea level rise. Results indicate that the total freshwater volume increases about
19% under the effects of future rainfall patterns. If moderate pumping is included, with rates
increasing at 1.76% to correspond with increasing population, the volume increases only by
12%. If just considering sea level rise, then the volume decreases by 14%. With aggressive
pumping, corresponding to an annual population growth rate of 9%, but no sea level rise, the
volume decreases by 24%. With aggressive pumping and sea level rise, the freshwater lens is
rapidly depleted.
This study quantifies the major future impacts on groundwater of the atoll islands in
Maldives. Similar methodologies using output from GCMs can be used for other atoll island
nations, such as the Republic of Marshall Islands, Federated States of Micronesia, and Gilbert
Islands. For the Maldives, results from this study can be used in conjunction with population
iii
growth estimates to determine the feasibility of including groundwater in water resources
planning and management for the country.
iv
TABLE OF CONTENTS
ABSTRACT .................................................................................................................................... ii
LIST OF TABLES ........................................................................................................................ vii
LIST OF FIGURES ....................................................................................................................... ix
CHAPTER 1: INTRODUCTION OF ATOLLS IN THE REPULIC OF MALDIVES .............. 1
1.1 Geography and Population of the Maldives .......................................................................... 1
1.2 Climate and Water Resources of the Maldives ..................................................................... 1
1.3 Geology of Atoll Islands in the Republic of Maldives.......................................................... 3
1.4 Freshwater Lens .................................................................................................................... 3
1.5 Water resources in the Republic of Maldives ....................................................................... 5
1.5.1 Water resources and usages in Maldives ........................................................................ 5
1.5.2 Threat to the freshwater lens in the Maldives................................................................. 7
1.6 Brief Historical review of estimating freshwater lenses ..................................................... 10
1.7 Objectives of this study ....................................................................................................... 11
CHAPTER 2: ASSESSING GOURNDWATER SUPPLY OF THE MALDIVES WITH 2-D
MODELING................................................................................................................................. 13
2.1 Introduction ......................................................................................................................... 13
2.2 Methodology ....................................................................................................................... 14
2.2.1 Construction of Island Subsurface Models using SUTRA ........................................... 15
2.2.2 Assigning Recharge Rates during 1998-2050 .............................................................. 18
2.2.3 Summary of Simulations .............................................................................................. 23
2.3 Results ................................................................................................................................. 23
2.3.1 Accepted GCMs for the Maldives Region ................................................................... 24
2.3.2 Lens Thickness Fluctuation and Trends through 2050 ................................................. 27
2.4 Summary and Concluding Remarks .................................................................................... 33
CHAPTER 3: ASSESSING IMPACTS OF RAINFALL PATTERNS, POPULATION
GROWTH, AND SEA LEVEL RISE ON GROUNDWATER SUPPLY IN THE REPUBLIC OF
MALDIVES USING 3-D MODEING.......................................................................................... 35
3.1 Introduction to Three-Dimensional modeling on atoll islands............................................ 35
3.2 Methods: .............................................................................................................................. 38
v
3.2.1 Gan of Laammu Atoll ................................................................................................... 38
3.2.2 Model development ...................................................................................................... 40
3.2.3 Model Calibration ......................................................................................................... 46
3.2.4 Estimating the Effect of Future Climate Scenarios on Gan’s Freshwater Lens ........... 47
3.3 Results ................................................................................................................................. 51
3.3.1 3-D views of simulation results .................................................................................... 51
3.3.2 Calibration Results ....................................................................................................... 53
3.3.3 Pumping effects ............................................................................................................ 54
3.3.4 Sea-level rise effects ..................................................................................................... 59
3.3.5 Combined effects of sea-level rise and aggressive pumping ........................................ 60
3.4 Discussion ........................................................................................................................... 60
3.4.1 Advantage and disadvantage of the model ................................................................... 60
3.4.2 Effects of changing rainfall pattern .............................................................................. 62
3.4.3 Effects of pumping ....................................................................................................... 62
CHAPTER 4: SUMMARY........................................................................................................... 65
REFERENCE ................................................................................................................................ 67
APPENDIX I ................................................................................................................................ 75
APPENDIX II ............................................................................................................................... 87
vi
LIST OF TABLES
Table 1. Lens thickness and sustainable yield for 19 islands in Maldives from groundwater
investigation .................................................................................................................................... 7
Table 2. Properties of the atoll island aquifer system, including general properties, and properties
for the Holocene and Pleistocene aquifer units (after Bailey et al., 2014a).................................. 17
Table 3. General Circulation Models (GCM) and their organizations. ........................................ 19
Table 4. Statistical criteria for evaluating GCMs. The weighting factor assigned to each criterion
also is shown. ................................................................................................................................ 20
Table 5. Model performance results for monthly rainfall rates in Region 1 and RCO Scenario 2.6,
ranking best to worst according to the total score......................................................................... 25
Table 6. Accepted GCMs for the three regions for RCP2.6 and for RCP8.5. .............................. 26
Table 7. Average lens thickness under the center of the island for each island width and
geographic region, across all accepted GCMs from the RCP2.6 and RCP8.5 scenarios. The first
set of values is averages through the years 2011-2030, and the second set is for the years 20312050. The GCM index corresponds to the order listed in Table 6. ............................................... 34
Table 8: Properties of the atoll island aquifer system, including general properties, and properties
for the Holocene and Pleistocene aquifer units for 3-D model construction ................................ 42
Table 9: All the GCMs used for simulations of each scenario. ................................................... 48
Table A1. Model performance results for monthly rainfall rates in Region 1 and RCP Scenario
4.5, ranking best to worst according to the total score.................................................................. 75
vii
Table A2. Model performance results for monthly rainfall rates in Region 1 and RCP Scenario
6.0, ranking best to worst according to the total score.................................................................. 76
Table A3. Model performance results for monthly rainfall rates in Region 1 and RCP Scenario
8.5, ranking best to worst according to the total score.................................................................. 77
Table A4. Model performance results for monthly rainfall rates in Region 2 and RCP Scenario
2.6, ranking best to worst according to the total score.................................................................. 78
Table A5. Model performance results for monthly rainfall rates in Region 2 and RCP Scenario
4.5, ranking best to worst according to the total score.................................................................. 79
Table A 6. Model performance results for monthly rainfall rates in Region 2 and RCP Scenario
6.0, ranking best to worst according to the total score.................................................................. 80
Table A7. Model performance results for monthly rainfall rates in Region 2 and RCP Scenario
8.5, ranking best to worst according to the total score.................................................................. 81
Table A8. Model performance results for monthly rainfall rates in Region 3 and RCP Scenario
2.6, ranking best to worst according to the total score.................................................................. 82
Table A 9. Model performance results for monthly rainfall rates in Region 3 and RCP Scenario
4.5, ranking best to worst according to the total score.................................................................. 83
Table A10. Model performance results for monthly rainfall rates in Region 3 and RCP Scenario
6.0, ranking best to worst according to the total score.................................................................. 84
Table A11. Model performance results for monthly rainfall rates in Region 3 and RCP Scenario
8.5, ranking best to worst according to the total score.................................................................. 85
Table A12. Average lens thickness under the center of the island for each island width and
geographic region, across all accepted GCMs from the RCP8.5. The first set of values is
averages through the years 2011-2030, and the second set is for the years 2031-2050. The GCM
index corresponds to the order listed in Table 6. .......................................................................... 86
viii
LIST OF FIGURES
Figure 1. (A) Maldives geographical position in the Indian Ocean; (B) the atolls of the Maldives,
divided into three geographic regions used in this study: Region 1 is the area above 5 o N, Region
2 is between 5oN and 0o and Region 3 is below 0o; (C) a close-up of an example atoll (Laamu
Atoll), showing the coral islands surrounding the lagoon. ............................................................. 2
Figure 2. Hydrogeological cross section of a typical atoll island in the Maldives region; ............. 4
Figure 3: (B) Model domain adopted for the SUTRA modeling simulations, with boundary
conditions and fluid flux boundary (recharge at at the water table). ............................................ 16
Figure 4. Bar chart of sorted ranking score for the three geographic regions for the RCP2.6
scenario. All GCMs are grouped by identified change points. ..................................................... 26
Figure 5. Comparison of time series plots and a PDF plot for an accepted GCM (CSIRO-Mk3-60) (A, B) and a rejected GCM (bcc-csm1-1) (C, D) in Region 1 for RCP2.6. ............................. 27
Figure 6. Time series plot of lens thickness for each accepted GCM in Region 1 for (A) the 600
m and 1100 m islands, and (B) the 200 m and 400 m islands. Solid lines in each series represent
the average values of all simulations for a given island size from both RCP 2.6 and RCP.8.5. The
dashed lines are results from using historical daily rainfall data in the SUTRA models from
1998-2011. .................................................................................................................................... 28
Figure 7. Salinity distribution in the island subsurface, with the contour colors corresponding to
the percent of salt in the groundwater related to the salt content of seawater. Dark blue
corresponds to freshwater. Red corresponds to seawater, with a mixing zone between. The top
two graphs show the freshwater lens during the (A) dry season and (B) wet season for the 200 m
island. The bottom two graphs show the freshwater lens during the (C) dry and (D) wet season
for the 600 m island. ..................................................................................................................... 30
Figure 8. Lens thickness PDF plot for 400 m and 1100 m islands in the three regions, for the
RCP2.6 scenario. ........................................................................................................................... 31
ix
Figure 9. Boxplots of freshwater lens thickness comparing simulation results from RCP 2.6 and
RCP 8.5 for the 600 m islands. The graphs on the left show results for the three regions using
scenario RCP 2.6; the graphs on the right show results for the three regions using scenario
RCP8.5. The GCM index corresponds to the order listed in Table 6 ........................................... 32
Figure 10. Freshwater lens thickness contour plot after groundwater investigation by Bangladesh
Consultants, Ltd (2010a, b, c, d). .................................................................................................. 37
Figure 11. Geographic location of island of Gan and its historical monthly rainfall ................... 39
Figure 12: Map of Gan and the location of three villages. ........................................................... 40
Figure 13. Top view and cross section view of the mesh of Gan. The island surface is the red
area in Graph A with elevation of 0 m. Other colors presents the elevation change of the ocean.
....................................................................................................................................................... 43
Figure 14. Graph A, B, C, D shows the process of making the mesh. In Graph A, the mesh is fine
everywhere and even more fine for island surface but it takes 2.3 minutes to run one time step.
Graph B decreased the size of the mesh but it still take about 1.7 minutes/ time step. In Graph C,
it remains the fine grid size near the coast but coarsen the grid far from coast. It takes about 1.3
minutes/ time step. In the last graph, it enlarged the ocean grids. The more far the grid from coast,
the coarser they are. The grids in the center are coarsened but the ones around coast line remain
fine. It takes about 1minutes/ time step. ....................................................................................... 44
Figure 15. Pumping area in the model (Graph B) that represents the pumping in Gan(Graph A).
Three green areas simulate the pumping in three villages in Gan. ............................................... 46
Figure 16. Time series plots of precipitation (Graph A) and corresponding recharge (Graph B)
from five selected GCMs with scenario Rcp2.6 from the year of 2013 to 2050 for island of Gan.
....................................................................................................................................................... 49
Figure 17. Estimated conservative and aggressive population growth and pumping rate in the
future for Gan ................................................................................................................................ 50
x
Figure 18. Land surface comparison before and after sea-level rise. The blue line is the contour
before sea-level rise whereas the red one is after sea-level rise. .................................................. 52
Figure 19. Three dimensional review of the SUTRA modeling results. The color represents the
salt concentration. The red color has highest salt concentration that represents seawater. The blue
color is freshwater. Other colors are the mixture of fresh and seawater....................................... 53
Figure 20. Observed vs. modeled lens thickness plot with different vertical hydraulic
conductivities. The
=
/ shows the best match. .............................................................. 54
Figure 21. Comparison of freshwater volume time series plots from the scenarios with pumping
and without pumping. ................................................................................................................... 55
Figure 22. Time series plot of freshwater volume plot from pumping. The loss is the freshwater
volume difference between pumping scenario and non-pumping scenario. ................................. 55
Figure 23. Time series plots of freshwater volume and lens thickness for all selected GCMs .... 57
Figure 24. Graph A and B show the time series plots of freshwater volume from scenario Rcp2.6
and Rcp8.5. Graph C show the time series plot of freshwater lens thickness. ............................. 59
Figure 25. Time series plot comparison of freshwater volume between simulations that before
sea-level rise and after sea-level rise. ........................................................................................... 60
Figure B1. Comparison of historical rainfall data time series plots from worst GCMs for RCP.2.6
....................................................................................................................................................... 87
Figure B2. Time series plot of lens thickness for each accepted GCM in Region 2 .................... 88
Figure B3. Time series plot of lens thickness for each accepted GCM in Region 3 .................... 88
Figure B4. Boxplots of best five GCMs for each scenario in each region for 200m island ......... 89
xi
Figure B5. Comparison of scenario 26 boxplot of fresh lens for different three regions for 200 m
island ............................................................................................................................................. 90
Figure B6. PDF plot of fresh lens between GCMs in 200 m island ............................................. 91
Figure B7. Boxplots of best five GCMs for each scenario in each region for 400m island ......... 92
Figure B8. Comparison of scenario 26 boxplot of fresh lens for different three regions for 400 m
island ............................................................................................................................................. 93
Figure B9. PDF plot of fresh lens between GCMs in 400 m island ............................................. 94
Figure B10. Boxplots of best five GCMs for each scenario in each region for 600m island ....... 95
Figure B11. Comparison of scenario 26 boxplot of fresh lens for different three regions for 600
m island ......................................................................................................................................... 96
Figure B12. PDF plot of fresh lens between GCMs in 600 m island ........................................... 97
Figure B13. Comparison of scenario 26 boxplot of fresh lens for different three regions for 1100
m island ......................................................................................................................................... 98
Figure B14. Boxplots of best five GCMs for each scenario in each region for 1100m island ..... 99
Figure B15. PDF plot of fresh lens between GCMs in 1100 m island ....................................... 100
Figure B16. Lens thickness PDF plot for 200m islands ............................................................. 101
Figure B17. Lens thickness plot for 600 m islands..................................................................... 102
xii
CHAPTER 1: INTRODUCTION OF ATOLLS IN THE REPULIC OF MALDIVES
This first chapter gives an introduction about atoll islands, atoll island hydrogeology, and
groundwater resource of the coral islands of the Republic of Maldives. The problem statement
regarding groundwater resources will be discussed, and objectives of the thesis will be outlined.
1.1 Geography and Population of the Maldives
The Republic of the Maldives (Figure 1) consists of approximately 2000 coral islands, each
located in one of 26 atolls within the region of 50S -100N latitude in the Indian Ocean
(Karthikheyan, 2010). An atoll (Figure 1C) is a circular chain of small islands and coral reef
surrounding a shallow lagoon, underpinned by carbonate platforms that extend downward to a
volcanic edifice. The central lagoon is the result of the outgrowth of the coral reef due to a
preference for sediment-free waters (Falkland and Custodio, 1991), and usually is shallow and
assumed to have the same salinity as the surrounding ocean (Terry and Ting, 2012). The total
land area of the Maldives is approximately 300 km2 (MEE, 2011), with many of the inhabited
islands having a land surface area of less than 1 km2. Island widths range between 100 m to
about 1200 m. Maximum ground surface elevation is 2.4 m above sea level, and 90% of the land
area has an elevation of less than 1 m. Approximately 200 islands are inhabited, with a total
population of 320,000 (The World Bank, 2011).
1.2 Climate and Water Resources of the Maldives
The Maldives experiences a warm and tropical climate year-round, with an average annual
temperature of 28.0 oC and an average relative humidity of 80%. The rate of rainfall increases
from north to south, with three distinctive regions of 50N -100N, 00 -50N, and 50S -00 having
1
average annual rainfall rates of 1715 mm, 1940 mm, and 2380 mm respectively, over the 19982011 period. A distinct dry season occurs during from January to April, particularly for the
northern regions. As the high permeability of islands soils precludes the formation of streams or
surface water bodies, typical of atoll islands (Urish, 1951), the communities rely on a
combination of water from rooftop catchment systems, desalinized water, and groundwater.
Figure 1. (A) Maldives geographical position in the Indian Ocean; (B) the atolls of the Maldives, divided into three
geographic regions used in this study: Region 1 is the area above 5 o N, Region 2 is between 5oN and 0o and Region 3
is below 0o; (C) a close-up of an example atoll (Laamu Atoll), showing the coral islands surrounding the lagoon.
2
1.3 Geology of Atoll Islands in the Republic of Maldives
The island’s aquifer usually consists of two layers, the Holocene aquifer resting on the
Pleistocene aquifer (Figure 2A). Many studies have found the existence of these two layers on
the atoll islands in the Indian Ocean before including Cocos Islands (Woodroffe and Falkland,
1997). During 2000 to 2001, Falkland did lots of coring on 16 islands in the Maldives; white
coral rock was found that might belong to the Pleistocene age. They also estimated that the
contact between the two layers, which is termed as “Thurber Discontinuity” (Thurber et al.,
1965), ranges from 9.5-24 m (Falkland, 2000, 2001). The contact is an important factor in
limiting the freshwater lens in large atoll islands because of the large difference between the
aquifer hydraulic conductivities (Hunt, 2007). The permeability (typically 5-10m/day) of upper
aquifer has been estimated to be one to two orders of magnitude less than that of the Pleistocene
aquifer(typically 50-1000m/day) (Falkland, 2000).
1.4 Freshwater Lens
Fresh groundwater is mostly contained in the Holocene aquifer. Rainfall that does not runoff
directly into the ocean and is not transpired by vegetation percolates through the shallow
unsaturated zone of the island and recharges the water table. The body of fresh groundwater in
the aquifer is referred to as a “freshwater lens” due to its geometrical shape (Figure 2), with the
maximum thickness of the lens typically occurring under the center of the island. Due to the
mounding of the water table and the resulting hydraulic gradient, the fresh groundwater flows
towards the perimeter of the island and discharges into the sea (Glover, 1964). Vertical
infiltration of precipitation expends the thickness of freshwater, which varies under a dynamic
equilibrium between moving freshwater and sea water. To simply quantify the thickness of the
3
freshwater lens, Badon and Herzberg used the difference in density between freshwater and
seawater to determine that the thickness of the freshwater lens is approximately forty times the
height of the freshwater level above sea level (Badon Ghyhen, 1889; Herzberg, 1901). However,
Ghyben-Herzberg principle is not always applicable because the interface of freshwater and
seawater is not a sharp sharp interface; instead, there is a mixing zone of freshwater and seawater,
which makes estimating the freshwater boundary very difficult. Various modeling methods used
to simulate the freshwater/seawater dynamics and the depth of this interface will be presented in
Chapters 2 and 3 of this thesis.
Figure 2. Hydrogeological cross section of a typical atoll island in the Maldives region;
The amount of fresh groundwater, which often is quantified by the metrics of freshwater lens
volume and maximum freshwater lens thickness, is controlled principally by the width of the
island, the permeability of the aquifer, the recharge rate to the water table resulting from
4
precipitation (Mather, 1975), and (if any) groundwater extraction through vertical wells or
horizontal infiltration galleries (White et al., 2007). Generally, freshwater lens volume and
thickness correlate positively with island size; for example, small coral islands (cross-section
width < 600 m) generally have thin (< 5 m) lenses, whereas larger coral islands (> 600 m)
generally have thick (5-20 m) lenses. As mentioned before, the input of freshwater recharge and
discharge to the sea establishes the dynamic equilibrium. A change in rainfall patterns can have a
significant impact on the volume of fresh groundwater in the aquifer. In addition, sea-level rise,
population growth and associated increase in groundwater extraction, and groundwater
contamination from leaching land surface pollution results in the freshwater lens being an
extremely fragile resource for small coral islands (Pernetta, 1992; Presley, 2005; Church et al.,
2006).
1.5 Water resources in the Republic of Maldives
This section summarizes all possible water resources used in Maldives and quantifies their
usages. It gives an overview of present situation and management of those water resources in
Maldives, followed by discussing all kinds of threats to freshwater lens of Maldives. One of the
objectives of this thesis is to evaluate some of the threats.
1.5.1 Water resources and usages in Maldives
Freshwater mainly comes from three resources in Maldives: rainwater, groundwater, and
desalinated seawater. The government’s goal is to provide 10 liters of safe water per person per
day for drinking and cooking purposes (MPHRE, 1998), with the majority of water coming from
rainwater catchment systems. Rainwater is usually collected into water tanks through a house
roof and gutter system. According to UNEP (United Nations Environment Programme), in 2005,
5
75% of the population collected rainwater from communal rainwater storage tanks or individual
household tanks (UNEP, 2005). Since 1994, the government has focused on providing 2914
high-density polyethylene (HDPE) tanks (capacity of 5000L) for community use on over 200
inhabited islands and implemented a program to provide household tanks (capacity of
1500~2500L) for each household in 2006 (WHO,2009). Besides rainwater, desalinated water is a
source for drinking water. The use of desalinated water is increasing, especially in the capital
Male. 35% of the population in Maldives has access to desalinated water with desalination plants
in 51 islands (MEE, 2011). However, after the Indian Ocean Tsunami of 2004, 30% of the
population had drinking water shortages (MPND, 2004). From the year of 2005 to 2011, More
than about 70 out of 200 inhabited islands reported water scarcity every year, ranging from 2.1
ML to 7.5 ML (MEE, 2011).
According to Beswick (2000), groundwater fulfills for non-potable usages such as sanitary
cleansing and toilet flushing, bathing and clothes washing. Estimates of total water usage are as
high as 175L/p/d (Beswick, 2000). However, after reviewing other studies, Falkland (2010)
recommended water usage as 120L/p/d, in which only 5-10L is from rain catchment, and the
bulk of water is from pumping. Groundwater is still the main domestic water resource for most
islands of the Maldives. There is still no adequate data for assessing fresh groundwater quantities
in all islands (MEE, 2011), although some groundwater investigations have been done for
nineteen islands (Falkland, 2000 2001; Bangladesh Consultants, Ltd., 2010a,b,c,d). The average
freshwater lens and sustainable yield for each of the studied island from these studies are listed in
Table 1. Adequate fresh groundwater resources exist in most islands. The data from these islands,
particularly the freshwater lens thickness, are used to test the model applications described in
Chapters 2 and 3.
6
Before 2005, Male Water and Sewerage Company (MWSC) were in charge of the promotion
of best practice water supply and sanitation provision for all the islands, especially providing
potable water to Male residents. The outer islands of Male managed their own water at the
household level. After July 2005, a new ministry, the Ministry of Environment, Energy and
Water (MEEW), share the authority with MWSC on national water sector management. MEEW
is now in charge of overall water policy for Maldives. However, the outer islands communities
are still responsible for the operation and maintenance of these new systems (GWP Consultants,
2006).
Table 1. Lens thickness and sustainable yield for 19 islands in Maldives from groundwater investigations (Falkland,
2000 2001; Bangladesh Consultants, Ltd., 2010a,b,c,d)
Atoll
Island Name
Addu Atoll
Haa Alifu Atoll
Haa Dhaalu Atoll
Gaafu Dhaalu
Atoll
Laamu Atoll
Noonu Atoll
Hithadhoo
Maradhoo
Feydhoo
Gan
Hoarafushi
Ihavandhoo
Dhidhdhoo
Kelaa
Filladhoo
Baarah
Hanimaadhoo
Nolhivaranfaru
Nolhivaram
Kulhudhuffushi
Kumundhoo
Thinadhoo
Area
(ha)
544
87
60
290
63
60
51
213
226
249
260
151
221
172
178
118
Average Freshwater Lens Thickness
(m)
8
2
6
15
0.5
3
2
8
0.5
6
4
0.5
4
7
8
5.5
Sustainable Yield
(kL/Day)
3,680
322
441
2,880
140
135
205
480
35
375
775
125
515
475
545
1000
Gan
Holhudhoo
Velidhoo
598
19.8
44.2
8.4
1.28
1.93
5600
100
240
1.5.2 Threat to the freshwater lens in the Maldives
Fresh groundwater can be a valuable water resource, however, it is very vulnerable. Sea-level
rising, changing rainfall pattern, shoreline erosion, sea water intrusion, pumping, and
groundwater contamination are all threats to freshwater lens.
7
Climate change is a huge concern for the Republic of Maldives. The rising sea-level and
seawater temperature are threats to the freshwater lens. The sea-level rise is predicted to rise a
half meter for the Maldives during 21st century (Woodworth, 2005), which will result in
significant shoreline recession. The general shoreline recession is 100 m corresponding to 1 m of
sea-level rise (1% slope) (Tysban et al., 1990). As a result, massive land areas will erode and
inundate due to the rising seawater; in turn, the thickness of the freshwater lens decreases as land
surface area shrinks (Volker et al. 1985). Another major concern from climate change is the
increase in water temperature. Coral bleaching would increase due to temperature increase in
lagoon water (Pernetta and Elder, 1990). Current death of corals is associated with thermal stress
(Pernetta, 1992). In 1998, 70-90% mortality of coral was reported in Maldives due to climate
related bleaching events (Rajasuriya et al., 2000). The reasons of this bleaching event, including
high temperatures, are discussed (Wilkinson et al., 1999). Global-mean land temperatures are
projected to increase about 5.5K by 2100 (Cox et al., 2000), which might make the bleaching
events more frequent and widespread. Death of coral reefs will accelerate the erosion of
shorelines and damage to the freshwater lenses.
Human activities influence shoreline erosion, in addition to climate changes. In the Maldives,
human activities like construction of causeways between islands, construction of wharves,
grouynes and breakwater, dredging, harbour works, sea defences and sand mining cause severe
erosion (Mörner et al., 2004; Rajasuriya et al., 2000; Richmond et al., 2006 ). Furthermore,
human activities tend to negatively impact the coral reefs. As mentioned before, as a country of
1192 coral islands with its great diversity and extent in corals, the reefs are essential for its
shoreline protection for Maldives (Rajasuriya et al., 2000). However, intensive coral mining has
caused dramatic reduction in coral varieties and abundance in some areas. For example, the
8
living coral reef in North Male is estimated to be exhausted within few decades (Brown and
Dunne, 1988). There is no complete evaluation of shoreline erosion for all atolls in Maldives,
however, with the evidence above, it is reasonable to conclude that the shorelines of the
Maldives are a severe risk of erosion.
Climate controls the rainfall patterns, so a change in climate results in more frequent
inordinate events, such as extreme rainfall, winds, storm surge tide and droughts (White et al.,
2007; Peinhardt, 2014; Sovacool, 2012; Hay, 2006). A long-term drought can deplete the
freshwater lenses since fresh groundwater continually discharges to the sea. During the dry
season of a normal year, from January to March in Maldives, the freshwater lens of small coral
islands almost depletes. In addition, surges of seawater during storms and typhoons can cause
damage to the freshwater lens. A storm surge is a high rise of seawater that erodes the freshwater
aquifer and overtops low-laying islands, causing seawater intrusion (Terry and Thaman 2008;
Spennemann 2006). It can take up to eleven months for a freshwater lens to recover from the
seawater intrusion (Terry and Falkland, 2010).
All of the above examples are natural stressors. However, anthropogenic stress, such as
population growth and groundwater contaminations, has already done a lot of damage to the
freshwater resources in Maldives (SOE, 2011; Pernetta, 1992). The relatively high-population
capital, Male, has run out of potable, fresh groundwater (Pernetta, 1992). Water scarcity occurs
often with significant population increase (SOE, 2004). The population of Maldives has tripled
in the last 35 years with an above average growth rate of 1.76% (World average: 1.17%) (SOE,
2011). High population growth creates a higher water demand via pumping.
9
Groundwater contamination is detected in sampled wells of all islands. Fecal coliform,
salinity, ammonia, nitrate and nitrite and phosphate are found at different levels for each island,
which renders fresh groundwater to be non-potable or highly unsafe for domestic use. From the
report “State of the Environment for Maldives” (2004), only 39 of the 198 islands’ groundwater
is suitable for drinking (MEE, 2004).
1.6 Brief Historical review of estimating freshwater lenses
There are several ways to estimate the freshwater lens thickness in atoll islands. A traditional
way of performing these estimates is through field measurements. Historically, the principal
means of measuring the thickness of the lens is through drilling monitoring wells and measuring
the salinity concentration of the groundwater through the depth of the profile. Recently,
geophysical methods were used (Falkland 2000, 2001) to estimate lens thickness and, if enough
measurements were taken, the volume of the lens. For example, during 2000 to 2001, Falkland
did salinity surveys on 10 islands in the Maldives using Electromagnetic (EM) surveying
(Falkland, 2000; Falkland, 2001). Falkland tested water salinity from several boreholes to
estimate the lens thickness with supplemental (EM) surveying. Thus, the relationship of EM
readings and lens thickness from salinity survey was developed and used to estimate the total
groundwater resources for each island. In 2009, Bangladesh Consultants performed groundwater
investigations on the islands of Gan, Thinadhoo, Holhudhoo and Velidhoo. They used the EM
method at more than 20 locations on each island to make contour maps and estimate the total
freshwater volume (Bangladesh Consultants, Ltd., 2010a,b,c,d). The advantage of this method is
that sufficient data can be obtained without drilling boreholes.
10
Another way of estimating lens thickness and lenses volume is through analytical and
numerical modeling, with these models solving mass balance equations for groundwater flow
and solute transport in coastal and island aquifers. Whereas analytical models can estimate lens
thickness while assume a sharp interface between freshwater and seawater in a homogeneous
aquifer (Fetter, 1972; Vacher, 1988; Bailey et al., 2013), numerical models can include spatiallyvarying aquifer properties, time-dependent fluid source terms (e.g. recharge to the freshwater
lens), horizontal and vertical flow patterns, and spatially-varying salt concentration in the
groundwater. Numerical models are able to assess the impact of drought, sea level rise, and
varying rainfall patterns on the dynamics of the freshwater lens.
1.7 Objectives of this study
In relation to the groundwater resources of the Republic of Maldives, the two main objectives
of this thesis are as follows:
1. Estimate the effects of future rainfall patterns on the freshwater lens for the islands of the
Maldives. Future rainfall patterns, which are necessary to compute future recharge
patterns for the islands, come from a set of General Circulation Models (GCMs) that
statistically compare favorably to historical rainfall patterns in the three geographic
regions of the Maldives.
2. Estimates the impacts of future rainfall, population and groundwater pumping growth,
and sea level rise on the volume and extent of the freshwater lens for a representative
island of the Maldives. The island selected is Gan island due to its moderate population
and potential for groundwater use.
11
A suite of 2D and 3D numerical models using the SUTRA (Saturated Unsaturated Transport)
modeling code and tested against field data from the Maldives will be used to accomplish these
objectives. The 2D modeling applications are summarized in Chapter 2, whereas the 3D model
application to the island of Gan is presented in Chapter 3. Summary and concluding remarks are
presented in Chapter 4.
12
CHAPTER 2: ASSESSING GOURNDWATER SUPPLY OF THE MALDIVES WITH 2-D
MODELING
This chapter illustrates how to use two-dimensional numerical modeling to assess
groundwater supply for different island sizes for Maldives. First, it gives a brief introduction and
review about 2-D modeling. Then, it shows how to use 2-D modeling to simulate densitydependent flow. Finally, the results give the prediction of lens thickness changes under climate
change from 2012 to 2050.
2.1 Introduction
Lam first started the numerical modeling for atoll islands by creating a model of pressure
field in an atoll using Darcy’s law (Lam, 1974). Then the model was developed for next decades
by several researchers (Lloyd et al., 1980; Falkland, 1983). Lloyd et al. developed the non-steady
solutions for the modeling for a Pacific Ocean atoll by finding out that steady state solutions
cannot be realistically applied (Lloyd et al. 1980). In 1984, Herman and Wheatcraft improved the
accuracy of the modeling by dividing the atoll aquifer into two aquifers that have large
differences in hydraulic conductivity (Herman and Wheatcraft, 1984). A finite-element code
SUTRA (Saturated and Unsaturated Transport) was developed by Voss (Voss, 1984) and has
been continuously used by many researchers for last 30 years on atoll islands (Hogan 1988;
Griggs 1989; Oberdorfer et al. 1990; Underwood et al. 1992; Griggs and Peterson 1993; Peterson
and Gingerich 1995; Gingerich and Voss, 2005; Bailey et al, 2009 ; Ketabchi et al. 2014 ).
SUTRA is also used for this study.
13
More recently, freshwater lens thickness was estimated for the islands of the Maldives during
the 1998-2011 time period (Bailey et al., 2014a) and during the coming decades under different
scenarios of sea level rise (Bailey et al., 2014b), with model results tested against the data from
Falkland (2000, 2001) and Bangladesh Consultants, Ltd. (2010a,b,c,d). Bailey et al. (2014a) used
code SUTRA (Voss and Provost, 2010) and daily recharge rates to simulate time-dependent lens
dynamics in a two-dimensional (2D) vertical cross-section of the islands, whereas Bailey et al.
(2014b) used an empirical model to calculate steady-state lens thickness according to island
width, aquifer hydraulic conductivity, depth to the contact between the Holocene and Pleistocene
aquifer units, and average annual recharge rate, with island width in future decades decreasing
depending on the rate of sea level rise. Results from both studies suggest that groundwater is a
viable option for water supply both now and in the future, with average lens thickness estimated
to be approximately 2.5 m for 400 m wide islands, 4.0 m for 600 m islands, and 12.0 m for 1100
m islands. Neither study, however, accounted for the effects of future rainfall patterns on
groundwater resources.
The objective of this study is to assess groundwater resources of the Republic of Maldives
under future climate conditions (through the year 2050) with a focus on the influence of future
variable rainfall patterns.
2.2 Methodology
This section presents the methodology for estimating freshwater lens thickness under future
rainfall patterns for the islands of the Maldives. Models are constructed for islands of varying
widths (200 m, 400 m, 600 m, 1100 m), with rainfall and recharge applied for the three different
geographic regions. Rainfall is obtained from GCMs that accurately replicate historical rainfall
patterns in the region of the Maldives. Whereas results will be analyzed in detail for 2011-2050,
14
simulations are run from 1998 so that results for 1998-2011 can be compared to results from
using historical daily rainfall data as a test of the GCM rainfall output.
2.2.1 Construction of Island Subsurface Models using SUTRA
Freshwater lens dynamics, i.e. freshwater-seawater interaction within the atoll island aquifer
system, were simulated using the SUTRA, which uses the finite element method to solve the
coupled density-dependent groundwater flow and solute transport governing equations. Required
parameters include porosity, permeability, specific yield, porous matrix compressibility, and
dispersivity. Model output includes nodal pressure, saturation, and solute concentration. Salt is
used as the solute in this study. Salt concentration of fresh groundwater is limited to 0.00089
kgsalt/kgwater. This corresponds to a chloride concentration of approximately 550 mg/L and an
overall salt concentration equal to 2.5% of the salt contained in seawater, which is a slightly
lower than the World Health Organization (WHO, 1972) recommended of 600mg/L. The body
of groundwater with salt concentrations less than this limit defines the freshwater lens.
Four different 2D finite-element meshes representing islands with widths of 200 m, 400 m,
600 m and 1100 m were constructed to represent the vertical cross-section of generic atoll
islands from lagoon side of the island to the ocean side (see Figure 3A). The outline of the model
domain is shown in Figure 3B for the model representing the 600 m island, with the top layer of
mesh nodes corresponding to mean sea level. Each mesh was finely discretized within the
immediate vicinity of the island subsurface, with particularly high refinement along the upper
model layers and within the approximate location of freshwater lens development (Bailey et al.,
2009). For example, node spacing in the top 4 m of the model is between 0.2 m and 0.3 m, then 1
m between 4 and 15 m below sea level, and then 0.75 m between 15 m and 25 m below sea level.
The 400 m island has 7,832 elements and 8,055 nodes and the 1100 m island has 18424 elements
15
and 18850 nodes, with the other models for 200 m and 600 m containing numbers of elements
and nodes commensurate with the island width.
Figure 3: (A) Typical island cross section for coral islands (B) Model domain adopted for the SUTRA
modeling simulations, with boundary conditions and fluid flux boundary (recharge at the water table).
Each island model received the same aquifer properties, assuming uniform geologic
formation across the region of the Maldives. This seems appropriate given the similarity between
estimated K values in the southern region (Falkland, 2000) and the northern region (Falkland,
2001). General aquifer properties (Table 2) include compressibility of the porous matrix, specific
16
yield, and longitudinal and transverse dispersivity for salt transport. The subsurface is divided
into an upper Holocene aquifer (13 m thick on ocean side, increasing to 18 m thick on the lagoon
side) and a lower Pleistocene aquifer (485 m thick). Holocene aquifer K (horizontal K and
vertical K equal to 75 m/d and 15 m/d, respectively) was determined in a previous study (Bailey
et al., 2014b) through comparing model simulation results with observed lens thickness values
for the 1998-2011 period. Results are shown in Supplementary Data. Due to Pleistocene aquifer
K estimated to be 1 to 2 orders of magnitude higher than Holocene aquifer K (Woodroffe and
Falkland, 1997), Pleistocene aquifer K is set to 5000 m/d. Water table storage was accounted for
by setting the nodal specific storage values to the specific yield divided by half of the element
vertical thickness (Griggs and Peterson, 1993). The nodes along the lagoon basement and reef
are assigned specified pressure based on the depth of each node below sea level, and a constant
specified concentration of 0.0357 kg/kg to represent seawater. The bottom of the mesh and the
boundary simulating the limit of the lagoon were assigned no-flow boundaries.
Table 2. Properties of the atoll island aquifer system, including general properties, and properties for the Holocene
and Pleistocene aquifer units (after Bailey et al., 2014a).
Parameters
Value
Units
Source
General Aquifer Properties
Compressibility of Porous
Matrix
Specific Yield
Longitudinal Dispersivity αL
Transverse Dispersivity αT
1.00 x 10-9
0.20
6.0
0.05
m2/N
m3/m3
m
m
Peterson and Gingerich 1995
Griggs and Peterson 1993
Holocene Aquifer
Holocene Porosity
Holocene thickness
Holocene horizontal K
Holocene vertical K
0.2
13 to 18
75
15
m3/m3
m
m/d
m/d
Anthony 1997
Hamlin and Anthony 1987
Calibrated value (this study)
Calibrated value (this study)
Pleistocene Aquifer
Pleistocene Porosity
Pleistocene thickness
0.3
485
m3/m3
m
Swartz 1962
17
Griggs and Peterson 1993
Pleistocene horizontal K
Pleistocene vertical K
5000
1000
m/d
m/d
Oberdorfer et al. 1990
Oberdorfer et al. 1990
2.2.2 Assigning Recharge Rates during 1998-2050
2.2.2.1 Overall procedure of using General Circulation Models
Recharge derived from rainfall is applied to the nodes along the top of the model (see Figure
3B), with spatially-uniform recharge rates assumed across the width of the island. Daily recharge
rates from 1998 to 2050 are calculated by statistically downscaling monthly output from GCMs
participating in the Coupled Model Intercomparison Project 5 (CMIP5) (Meehl et al., 2009;
Taylor et al., 2012). Rainfall output from both the lowest emission scenario [Representation
Concentration Pathway (RCP) 2.6, corresponding to a limited radiative forcing of 2.6 W/m2 by
the year 2100] and the highest emission scenario RCP8.5 are used to provide end-member
climate conditions. We acknowledge the World Climate Research Programme's Working Group
on Coupled Modelling, which is responsible for CMIP, and we thank the modeling groups (Table
3) for producing and making available their model output. 24 GCMs are used in total.
The overall process of calculating daily recharge is as follows:
1) Retrieve monthly rainfall depths as simulated by the GCMs
2) Statistically compare patterns of rainfall between the GCMs and historical monthly
rainfall depths over the 1998-2011 time period, to either accept or reject each GCM
3) For the accepted GCMs, downscale the monthly rainfall depths to daily rainfall depths
using a Markov chain algorithm
4) Use daily rainfall depths to calculate daily recharge using a soil water balance model
(Falkland, 1994).
18
The process is performed for both RCP2.6 and RCP8.5 and for each of the three geographic
regions of the Maldives: Region 1 (50N-100N, 700E-750E), Region 2(00-50N, 700E-750E), and
Region 3(50S-00, 700E-750E). Steps 2), 3), and 4) will now be described.
Table 3. General Circulation Models (GCM) and their organizations.
Modeling Center (or Group)
Institute ID
Model Name
Model
Index
Beijing Climate Center, China Meteorological Administration
BCC
BCC-CSM1.1
1
National Center for Atmospheric Research
NCAR
CCSM4
2
Community Earth System Model Contributors
NSF-DOE-NCAR
CESM1(CAM5)
3
Commonwealth Scientific and Industrial Research Organization in
collaboration with Queensland Climate Change Centre of Excellence
CSIRO-QCCCE
CSIRO-Mk3.6.0
4
The First Institute of Oceanography, SOA, China
FIO
FIO-ESM
5
GFDL-CM3
6
GFDL-ESM2G
7
GFDL-ESM2M
8
GISS-E2-H-p1
9
GISS-E2-H-p2
10
GISS-E2-h-p3
11
GISS-E2-H-p1
12
GISS-E2-H-p2
13
GISS-E2-H-p3
14
HadGEM2-AO
15
NOAA Geophysical Fluid Dynamics Laboratory
NOAA GFDL
NASA Goddard Institute for Space Studies
NASA GISS
National Institute of Meteorological Research/Korea Meteorological
Administration
NIMR/KMA
Met Office Hadley Centre (additional HadGEM2-ES realizations
contributed by Instituto Nacional de Pesquisas Espaciais)
MOHC
(additional
realizations by
INPE)
Institut Pierre-Simon Laplace
IPSL
Atmosphere and Ocean Research Institute (The University of Tokyo),
National Institute for Environmental Studies, and Japan Agency for
Marine-Earth Science and Technology
Japan Agency for Marine-Earth Science and Technology, Atmosphere
19
16
HadGEM2-ES
17
IPSL-CM5A-LR
18
IPSL-CM5A-MR
19
MIROC
MIROC5
20
MIROC
MIROC-ESM
21
and Ocean Research Institute (The University of Tokyo), and National
Institute for Environmental Studies
Norwegian Climate Centre
MIROC-ESMCHEM
22
NorESM1-M
23
NorESM1-ME
24
NCC
2.2.2.2 Statistical Comparison between GCMs and Historical Rainfall
Seven statistical criteria (Table 4) are used to assess the performance of each GCM (Fu et al.,
2013). To eliminate bias, the overall performance is assessed using the combined effects of all
criteria. Each individual criterion is associated with weighting factors, with some having weights
of 0.5 because they are complementary to another criterion. The total rank score of a GCM is
calculated as follows, with low scores indicating a closer agreement between the GCM output
and historical data (RC represents Ranking Score):
RCtotal  RCmean RE  RCStd RE  RCNRMSE  RCCorr   0.5  RCBS    0.5  RCS score   RCKendal Slope (1)
Table 4. Statistical criteria for evaluating GCMs. The weighting factor assigned to each criterion also is shown.
Statistic Criterions
Formula
Weighting factor
umodel  uob.
uob.
Mean Relative Error (Mean RE)
Std model  Stdob.
Stdob.
Standard Deviation Relative Error (Std RE)
1 n
 ( X mi  X oi )2
n i 1
1 n
 ( X Oi  X O )2
n i 1
Normalized Root Mean Square Error (NRMSE)
1.0
1.0
1.0
Correlation Coefficient (Corr)
Calculation in Matlab
1.0
Brier Score (BS)
1 n
( Pmi  Poi )2
n i 1
0.5
Skill Score
Minimum  P
n
i 1
mi
20
, Poi 
0.5
Kendall Slope
Calculated in Pro-UCL
1.0
Monthly mean relative error (mean RE) and standard deviation relative error (Std RE) are
used to quantify the similarity between modeled and historical rainfall depths; the Normalized
Root Mean Square Error (NRMSE) is used to compare the similarity of two time series by
considering both mean value and standard deviation; the Correlation Coefficient (Corr) is to
evaluate both the annual cycle and the spatial distribution of monthly rainfall; the Brier Score
(BS) and Skill Score (Sscore) are used to evaluate the GCM probability density functions (PDFs)
of monthly rainfall; and the Mann-Kendall Slope determines the magnitude of long-term trends
of time series. The resulting total scores (RCtotal) for each GCM are ranked and sorted, and the
moving range, i.e. the difference in scores between two successive GCMs, is used to detect
change points (Fu et al., 2013). In this study, any substantial increase in score between
successive GCMs constituted a change point, with the group of GCMs having a score lower than
the change point accepted for use in calculating recharge, and the remaining GCMs rejected.
This process is performed for both RCPs (RCP2.6 and RCP8.5) and for each of the three
geographic regions.
2.2.3 Statistical Downscaling to Daily Rainfall Depths
To obtain daily precipitation depths, the monthly data from the accepted GCMs are
downscaled statistically according to patterns of historical daily rainfall depths. A Markov chain
algorithm (Todorovic and Woolhiser, 1975; Srikanthan and McMahon, 2001) generates daily
wet/dry sequences for each month of historical data, fits shape parameters for gamma
distributions that describes the classifications, and then uses these distributions to provide
rainfall depths for future wet days. The Gamma distribution is used since it is a common
21
statistical model for days with non-zero precipitation (Coe and Stern, 1982; Srikanthan, 2005).
Simulated daily rainfall depths are scaled so that monthly sums equal monthly rainfall depths
from the GCM output. The algorithm is expressed by:
X t | X t 1  Markov  P , p1 
(2)
where P is the transitional probability matrix whose elements pij are defined by:
pij  Pr( X t  i | X t 1  j )
i,j = wet or dry
(3)
and p1 is the probability distribution vector of the wet/dry classifications (Srikanthan and
McMahon, 2001). Classification (wet/dry) for each historical month is determined by comparing
the monthly rainfall depth with the median of average rainfall across all years of data, after
which each month of the GCM simulation time also is assigned a wet/dry classification. The
wet/dry conditions of each day are determined using the same methodology and applied to the
days of each future month, with values drawn from Gamma distributions created for each
monthly of the year.
2.2.2.3 Calculating Daily Recharge from Daily Rainfall Depths
Daily recharge to the freshwater lens is estimated using the downscaled daily rainfall depths
and the daily soil water balance model of Falkland (1994), which accounts for canopy
interception, evapotranspiration (ET), soil field capacity, and soil wilting point, with recharge to
the water table occurring when soil water storage exceeds soil field capacity. On days when soil
water storage is below field capacity, ET is limited to 20% of the potential value (Lloyd et al.,
1980). The soil water balance equation (Falkland, 1994) is described as below. It does not count
for surface runoff which does not usually happen because of high infiltration capacity of the
coral soils. The recharge to the water table only happens when soil moisture content exceeds
field capacity. Precipitation is the daily value from downscaled monthly data. Evaporation is
22
usually estimated from using both Penman and Pan Methods. Transpiration rates is estimated
from measurement of coconut tree transpiration rate which is about 400mm-750 mm/year/tree.
R (mm) is the daily recharge
� = � − ��� + �
(4)
P (mm/day) is the daily rainfall
��� (mm/day) is the actual evaporation from all surfaces which included evaporation from
interception storage
� mm is the change in storage within the soil moisture zone. Maximum and minimum
limits are set for soil moisture to calculate the storage change
As with previous atoll modeling studies in the Federated States of Micronesia (Bailey et al.,
2009) and the Maldives (Bailey et al., 2014a), interception depth is set to 1.0 mm, field capacity
and wilting point are set to a soil water content of 0.15 and 0.05, respectively, and potential ET is
set to a constant daily value of 3.5 mm/day, which is the average value of ET in Maldives.
2.2.3 Summary of Simulations
Simulations are run for each accepted GCM for both RCPs (RCP2.6, RCP8.5), for each of
the four island width models and for each of the three geographic regions. The initial conditions
(salinity concentration and pressure at each mesh node) for each 1998-2050 simulation are
achieved by imposing a steady recharge rate until the freshwater lens reaches steady-state
conditions, followed by a transient simulation from 1991-1998 using recharge calculated from
historical daily rainfall rates. Results of each model simulation are processed to determine the
thickness of the freshwater lens under the center of the island.
2.3 Results
23
2.3.1 Accepted GCMs for the Maldives Region
Table 5 shows the statistical scores and total score for each GCM for Region 1 and RCP 2.6.
Tables summarizing GCM scores for each region for both the RCP 2.6 and RCP 8.5 scenarios
are contained in Supplementary Data. RCmean RE ranges from 0.5% to 58%, with most GCMs
having a small value. However, most RCNRMSE values are larger than 1.0, signifying a poor match
with the historical rainfall series in terms of both mean value and standard deviation. RCCorr
ranges from 0.3 to 0.7, with MIROC5 having the highest value. RCBS and RCS score values range
from 3.3 to 16.6 and from 85 to 110, respectively, with NorESM1-ME having the best match (i.e.
lowest RCBS and highest RCS score ).
The GCM scores for the three geographic regions for RCP2.6 are plotted in Figure 4 in
ascending order, with the GCMs divided into groups according to identified change points. For
each region there are three groups, with the GCMs within Group 1 accepted for use in the
groundwater modeling simulations. The accepted GCMs each of the three regions and for both
RCP scenarios are listed in Table 6. In general, there are more accepted GCMs in Region 1 than
the other regions, and in Region 2 than in Region 3. Several GCMs are consistently accepted in
most or all regions, such as CESM1-CAM5 and MIROC5. To demonstrate the difference
between accepted and rejected GCMs for a particular region, Figure 5 compares the times series
and PDF of historical and GCM rainfall data, with an accepted model (CSIRO-Mk3-6-0) shown
in Figure 5a,b and a rejected model (bcc-csm1-1) shown in Figure 5c,d. The time series of
historical rainfall shows a strong wet season – dry season pattern, with magnitude of rainfall rate
approximately uniform across all years. CSIRO-Mk3-6-0 matches the historical pattern very well,
whereas bcc-csm1-1 greatly over-predicts the rainfall in the region. These trends can also be seen
in the accompanying PDFs.
24
Table 5. Model performance results for monthly rainfall rates in Region 1 and RCO Scenario 2.6, ranking best to
worst according to the total score.
GCM
Mean RE(mm)
Std RE
NRMSE
Corr
BS
S score
Kendal Slope (mm/year)
Total Score
MIROC5
-0.07
0.05
0.85
0.66
4.11
103
0.0032
28
GFDL-CM3
0.09
0.17
0.56
0.56
5.03
109
0.003
43
CSIRO-Mk3-6-0
-0.12
-0.07
0.87
0.61
6.03
88
0.0011
47.5
IPSL-CM5A-MR
-0.01
0.90
0.90
0.56
4.41
99
0.003
50
GFDL-ESM2M
-0.03
0.05
1.01
0.51
3.34
102
0.009
51.5
FIO-ESM
0.07
0.04
1.12
0.40
3.81
99
0.0024
53
NorESM1-M
-0.03
0.18
1.12
0.48
6.59
98
0.0013
55.5
CESM1-CAM5
0.26
0.01
1.01
0.56
4.81
92
0.0019
56
HadGEM2-ES
-0.19
0.00
1.06
0.47
6.26
95
0.0026
63
CCSM4
0.24
0.10
1.08
0.53
4.53
93
0.0044
73
NorESM1-ME
0.01
0.18
1.28
0.32
3.32
110
0.005
73
GISS-E2-H p3
-0.05
0.33
1.23
0.47
4.67
107
0.0054
75.5
GFDL-ESM2G
-0.22
-0.27
0.99
0.44
4.13
98
0.0041
76
GISS-E2-R p3
-0.09
0.23
1.19
0.45
5.03
91
-0.0003
77
IPSL-CM5A-LR
-0.16
-0.44
0.87
0.54
5.21
87
0.0037
77.5
GISS-E2-R p2
-0.11
0.29
1.19
0.49
6.09
93
-0.0019
81.5
MIROC-ESM
0.54
0.45
1.38
0.62
4.41
100
0.0034
87.5
MIROC-ESM-CHEM
0.58
0.35
1.36
0.63
4.19
94
0.0047
93
MRI-CGCM3
-0.21
0.26
1.27
0.43
16.58
85
0.0005
93
GISS-E2-H p2
-0.06
0.46
1.33
0.47
5.93
92
0.0046
96.5
GISS-E2-R p1
-0.09
0.32
1.28
0.42
5.25
97
-0.0036
98
GISS-E2-H p1
-0.09
0.38
1.29
0.45
7.81
97
-0.0033
101
HadGEM2-AO
-0.23
0.11
1.25
0.35
8.81
89
0.0067
108.5
bcc-csm1-1
0.25
0.86
1.84
0.32
9.71
92
0.0091
135
25
Figure 4. Bar chart A, B, C of sorted ranking score for the three geographic regions for the RCP2.6 scenario. All
GCMs are grouped by identified change points.
Table 6. Accepted GCMs for the three regions for RCP2.6 and for RCP8.5.
Region 1
Scenario 2.6
Scenario 8.5
MIROC5
IPSL-CM5A-MR
GFDL-CM3
GFDL-ESM2M
CSIRO-Mk3-6-0
MIROC5
IPSL-CM5A-MR
CSIRO-Mk3-6-0
GFDL-ESM2M
CESM1-CAM5
NorESM1-M
GFDL-CM3
CESM1-CAM5
GFDL-ESM2G
HadGEM2-ES
CESM1-CAM5
Region 2
CESM1-CAM5
MIROC5
MIROC5
CCSM4
IPSL-CM5A-LR
IPSL-CM5A-MR
CCSM4
GISS-E2-R p1
GFDL-ESM2G
GFDL-ESM2G
GFDL-CM3
Region 3
CESM1-CAM5
CESM1-CAM5
CCSM4
MIROC-ESM
MIROC-ESM
CCSM4
26
Figure 5. Comparison of time series plots and a PDF plot for an accepted GCM (CSIRO-Mk3-6-0) (A, B) and a
rejected GCM (bcc-csm1-1) (C, D) in Region 1 for RCP2.6.
2.3.2 Lens Thickness Fluctuation and Trends through 2050
The time series of simulated monthly lens thickness for each accepted GCM for both RCP2.6
and RCP5 is shown in Figure 6 for each island width within Region 1. Similar plots can be made
for Regions 2 and 3. Each GCM is depicted by a light gray line, with the average monthly value
shown with a dark line. Model results using historical daily rainfall depths are shown in dotted
lines from 1998-2011, demonstrating the close match between using historical data and the
average of the accepted GCMs. Results show that larger islands (e.g. 600 m and 1100 m islands)
have much larger freshwater lenses than smaller islands (200 m and 400 m), with overall average
27
thickness of 0.3 m, 2.3 m, 5.6 m, and 11.7 m for 200 m, 400 m, 600 m, and 1100 m islands,
respectively. Overall, the islands with width of 200 m and 1100 m have a smaller scatter than
400 m and 600 m islands. The standard deviation of average lens thicknesses across the GCMs is
0.066 m, 0.22 m, 0.40 m and 0.19 m for 200m, 400m, and 600m and 1100m islands, respectively.
Figure 6. Time series plot of lens thickness for each accepted GCM in Region 1 for (A) the 600 m and 1100 m
islands, and (B) the 200 m and 400 m islands. Solid lines in each series represent the average values of all
simulations for a given island size from both RCP 2.6 and RCP.8.5. The dashed lines are results from using
historical daily rainfall data in the SUTRA models from 1998-2011.
28
Average simulated lens thickness for each simulation is shown in Table 7. Values are shown
for each island width within each of the three regions (R1, R2, R3), and for each accepted GCM.
The GCM index corresponds to the order listed in Table 6. Lens thickness for each island width
and for each region is averaged across the accepted GCMs for two time periods: the first 20
years (2011-2030) and the second 20 years (2031-2050). For example, the average lens thickness
for a 400 m island in region 2 (R2) is 2.81 m during 2011-2030 and 2.84 m during 2031-2050,
whereas the average lens thickness for an 1100 m island in R2 is 12.16 m during 2011-2030 and
12.13 m during 2031-2050.
The time series plot of lens thickness in Figure 6 shows that the freshwater lens responds
quickly to the distinct wet/dry season pattern of rainfall. The average lens thickness during the
wet season considering each accepted GCM for different size islands in Region I are 0.71 m, 3.1
m, 6.46 m and 12.73 m for 200 m, 400 m, 600 m, and 1100 m islands, respectively. The average
lens thickness during the dry season is 0.0 m , 1.4 m, 4.5 m and 10.57 m, corresponding to
decreases of 100%, 55%, 30%, and 17%, for the four island sizes, demonstrating that the effect
of the dry season is less for larger islands. Salt distribution in the subsurface during wet and dry
seasons is shown in Figure 7 for a 200 m island (a,b) and for a 600 m island (c,d). Blue
represents portions of the aquifer with a salt concentration less than the freshwater limit (2.5% of
the salt in seawater), and therefore the freshwater lens, and red represent portions of the aquifer
with seawater in the pores, with a mixing zone that grades from freshwater to seawater. The lens
of the 200 m island during the wet season (Figure 7a) is much thicker (0.71 m) than during the
dry season (0.003 m). Similar results occur for the 600 m island, comparing the lens during the
wet season (Figure 7c, lens thickness: 6.5 m) with the lens during the dry season (Figure 7d, lens
thickness: 4.4 m).
29
Figure 7. Salinity distribution in the island subsurface, with the contour colors corresponding to the percent of salt
in the groundwater related to the salt content of seawater. Dark blue corresponds to freshwater. Red corresponds to
seawater, with a mixing zone between. The top two graphs show the freshwater lens during the (A) dry season and
(B) wet season for the 200 m island. The bottom two graphs show the freshwater lens during the (C) dry and (D) wet
season for the 600 m island.
The difference in groundwater resources between the three geographic regions of the
Maldives is demonstrated in PDF plots of lens thickness (Figure 8). The plots for 400 m islands
in the three regions for RCP2.6 are shown in the upper plots (a,b,c), and the plots for 1100 m
islands are shown in the lower plots (d,e,f). The freshwater lens is thickest in the southern region
(R3: 5-10S) followed by the central region (R2: 0-5S), which is in agreement with the rainfall
patterns of the Maldives’ region with rates increasing north to south. Similar patterns (not shown)
occur for the 200 m and 600 m islands. The bi-modal shape of the distribution for the northern
region (Figure 8a,d) denotes the sharp contrast in rainfall and associated lens thickness between
the wet season and dry season. The southern regions do not experience as pronounced of a dry
season, and hence the distribution is uni-modal (Figure 8c, f).
30
Figure 8. Lens thickness PDF plot for 400 m and 1100 m islands in the three regions, for the RCP2.6 scenario.
31
The difference between the RCP2.6 and RCP8.5 scenarios is shown in Figure 9 for 600 m
islands in each of the three regions. The mean value of maximum lens thickness for RCP2.6 and
RCP8.5 for all three regions is 5.45 m and 5.43 m, 6.39 m and 6.28 m, 8.75 m and 8.81 m,
respectively. Average standard deviations of all GCMs in each of these two scenarios are 0.93
and 0.97, 0.90 and 1.02, 0.94 and 0.86, signifying the close agreement between the two pathway
scenarios.
Figure 9. Boxplots of freshwater lens thickness comparing simulation results from RCP 2.6 and RCP 8.5 for the
600 m islands. The graphs on the left show results for the three regions using scenario RCP 2.6; the graphs on the
right show results for the three regions using scenario RCP8.5. The GCM index corresponds to the order listed in
Table 6
In terms of the effects of changing rainfall patterns on groundwater resources, the results
shown in Table 7 indicate an increase in lens thickness between the first 20 years of the study
32
period (2011-2030) and the second 20 years (2031-2050). The majority of average values in the
second 20 years are larger than during the first 20 years, with the only exception being 1100 m
islands in region 2 (12.16 m in 2011-2030 compared to 12.13 m in 2031-2050). Overall average
increase in lens thickness is 3.8%, with 7.4%, 4.2%, 2.9%, and 0.6% for 200 m, 400 m, 600 m,
and 1100 m islands, showing the higher sensitivity of small islands to changes in temporal
rainfall and recharge patterns. These results suggest a slow natural increase in lens thickness, and
hence groundwater resources, over the next 40 years. This increase of course can be tempered by
anthropogenic influences such as groundwater pumping. Also, groundwater could become
contaminated in the short-term or long-term due to surface contamination or overwash events. In
general, however, results can be used by country officials in water resources management
decisions over the coming decades.
2.4 Summary and Concluding Remarks
This study provides an assessment of groundwater resources considering only changing
rainfall patterns, and does not consider the exploitation of these resources from a growing
population. Current population growth is estimated to be approximately 1.76 % in the Maldives,
which could impose a significant burden on groundwater resources. Three-dimensional modeling
approaches using estimates of population growth and water demand are needed to quantify the
maximum use of groundwater supply in future decades. Furthermore, groundwater quality must
be considered when assessing potable groundwater supply. These aspects can be conducted in
future studies, but likely not at the broad geographic scale assessed in this current study.
33
Table 7. Average lens thickness under the center of the island for each island width and geographic region, across
all accepted GCMs from the RCP2.6 and RCP8.5 scenarios. The first set of values is averages through the years
2011-2030, and the second set is for the years 2031-2050. The GCM index corresponds to the order listed in Table 6.
Average Lens
Island Width (m)
200
Last 20 Years (2031-2050)
First 20 Years (2011-2030)
Thickness (m)
400
600
1100
R1
R2
R3
R1
R2
R3
R1
R2
R3
R1
R2
R3
GCM 1
0.22
0.24
0.91
1.96
2.39
3.44
6.00
5.74
8.75
11.29
11.87
12.99
GCM 2
0.27
0.30
0.87
2.11
2.56
4.10
5.72
6.11
9.27
11.49
12.00
13.05
GCM 3
0.29
0.41
0.80
2.15
2.82
4.11
5.65
6.60
9.02
11.54
12.15
13.10
GCM 4
0.21
0.42
2.17
2.98
5.48
6.85
11.57
12.34
GCM 5
0.19
0.57
2.18
3.28
5.37
7.37
11.61
12.43
GCM 6
0.25
2.29
5.32
11.64
GCM 7
0.21
2.39
5.20
11.69
GCM 8
0.37
2.54
5.03
11.79
Mean
0.25
0.39
0.86
2.23
2.81
3.88
5.47
6.53
9.01
11.58
12.16
13.01
R1
R2
R3
R1
R2
R3
R1
R2
R3
R1
R2
R3
GCM 1
0.25
0.27
0.95
1.95
2.41
3.84
6.37
5.77
8.69
11.29
11.83
13.00
GCM 2
0.38
0.36
0.78
2.22
2.75
3.95
6.50
6.47
9.17
11.57
12.14
13.11
GCM 3
0.40
0.43
0.92
2.23
2.77
4.28
4.95
6.55
9.41
11.62
12.16
13.14
GCM 4
0.25
0.38
2.24
3.05
5.62
7.07
11.64
12.39
GCM 5
0.19
0.51
2.32
3.21
5.47
7.36
11.72
12.40
GCM 6
0.26
2.73
5.46
11.99
GCM 7
0.23
2.73
6.70
12.01
GCM 8
0.46
2.90
5.43
12.06
Mean
0.30
0.39
0.88
2.41
2.84
4.02
34
5.81
6.64
9.09
11.74
12.13
13.09
CHAPTER 3: ASSESSING IMPACTS OF RAINFALL PATTERNS, POPULATION
GROWTH, AND SEA LEVEL RISE ON GROUNDWATER SUPPLY IN THE REPUBLIC OF
MALDIVES USING 3-D MODEING
This chapter assesses the groundwater resources in one of the islands in Maldives by using
three dimensional modeling. It also examines the effects of changing rainfall pattern, increasing
pumping demand and rising sea level on the groundwater resources.
3.1 Introduction to Three-Dimensional modeling on atoll islands
To more accurately simulate the dynamics of freshwater-seawater interactions in island and
coastal aquifers, three-dimensional models have been used in groundwater modeling during the
past years. Compared to 2D vertical cross-section modeling, which was the method used in the
study detailed in Chapter 2, three dimensional models have the advantage of being able to
specify natural boundary conditions and the actual shape of the coast or island surface area,
include 3D spatial variability of material properties, and represent pumping in three dimensions
(Ghassemi et al., 1996; Ghassemi et al., 2000). Pumping can be represented in 2D models, but
major simplifications and assumptions must be employed since the actual cone of depression
cannot be simulated.
Several studies have used 3D modeling techniques to assess fresh groundwater supplies on
coral islands during the past two decades, using the modeling codes SUTRA (Voss and Provost,
2003;), HST3D (Kipp, 1987) SALTFLOW (Molson and Frind, 1994), and SEAWAT (Langevin
et al., 2007). Each has the capability to simulate density-dependent groundwater flow. A HST3D
model was built for Nauru Island in the Central Pacific Ocean to simulate the freshwater lens
35
(Ghassemi et al., 1996), although it is found to be inefficient in dealing with large islands and
fine grids even with the advanced version (Meurant, 1990). The SALTFLOW modeling code was
used for the aquifer on Home Island in the Indian Ocean (Ghassemi et al., 1998), although the
calibration was not very successful due to inadequate data for groundwater quality and quantity.
SEAWAT was used to simulate freshwater responses to climate change for a coral island
named Grande Glorieuse in Western Indian Ocean (Comte et al., 2014). A 3-D model with dual
aquifer system is created with total 67,830 cells to simulate the groundwater flow in a 14.7
island. 0.35 m of sea-level rise with a 0.64
loss of land area was considered in the model.
Rainfall and potential evaporation is projected to increase 3% and 6%, respectively. It also
evaluated the combination effect of climate change and vegetation on groundwater salinity.
However, a different groundwater modeling tool SUTRA is used in this thesis. This 3-D model
is created with seven times more cells for smaller island (6
). GCMs (General Circulation
Models) with different scenarios predict future rainfall and temperature. 0.5 m of sea-level rise is
projected by the end of 21st century with 12% land area loss. Besides, this thesis includes
pumping effects. SUTRA has been used successfully to simulate three-dimensional variabledensity flow in a coastal aquifer in southern Oahu in Hawaii by Gingerich and Voss (2005).
In December of 2009, the Bangladesh Company performed a groundwater investigation
survey in Gan using the EM method at more than 20 locations on each island. Three best-fit
mathematical relationships used in the past were applied to translate the EM readings to lens
thickness. They were used to provide contours of lens thickness (Figure 10) and a total volume
of fresh groundwater. The total volume of fresh groundwater at the time of the study was
estimated to be 14.2 million
, with an average freshwater lens thickness of 8.4 m (Bangladesh
Consultants, Ltd., 2010a,b,c,d).
36
Figure 10. Freshwater lens thickness contour plot after groundwater investigation by Bangladesh Consultants, Ltd
(2010a, b, c, d).
The main objective of this study is to assess fresh groundwater quantity of a major coral
island of the Maldives under future climate threats. The island chosen for the assessment is Gan
Island, located on Laammu atoll. Fresh groundwater lens thickness and volume are estimated for
Gan from 2012 to 2050. This study examines the effects of changing rainfall patterns, increasing
pumping needs and sea-level rise on lens thickness and volume. The SUTRA modeling code
(Voss and Provost, 2010) was selected as the tool to quantify lens dynamics in the 3D subsurface
37
of Gan Island during the 2012-2050 time period due to its recent successful application to coral
island aquifer systems. As compared to other full-island modeling studies, this study is the first
to quantify the effect of a growing population on the groundwater resources of the aquifer in a
changing climate. Methods of this study can be used for other oceanic islands, and results can be
valuable to the government of Gan Island specifically and to the government of the Maldives
generally for better water resources management.
3.2 Methods:
3.2.1 Gan of Laammu Atoll
As the biggest island in the Maldives, Gan in the Laammu atoll has a land area of 598 ha
with length of 7.8 km and width of 3.4 km (GoM, 2010). It is located in the eastern side of
Laamu atoll with latitude °55′and longitude7 °
′′
′. Gan’s population has grown rapidly
from 2,537 persons (AFD, 2007) in 2005 to 4,208 persons in 2009. The population for each of its
three villages (Thundi, Mathimaradhoo and Mukurimagu) were 2539, 748 and 921 (Bangladesh,
2010) (Figure 12). Gan falls into geographic Region II (see Chapter 1). Tropical Rainfall
Measuring Mission (TRMM) provides fourteen years of historical daily rainfall data from 1998
to 2011. Figure 11 displays the time series plot. The average annual rainfall is 1930 mm/year.
Wet and dry season are pronounced. High rainfall and very dry events are found around 2007.
38
Figure 11. Geographic location of island of Gan and its historical monthly average rainfall
39
Figure 12: Map of Gan and the location of three villages.
3.2.2 Model development
3.2.1.1 Model Construction
This study uses ModelMuse (Winston, 2009)—a graphical user interface for constructing
numerical models for USGS-based codes such as MODFLOW and SUTRA, is used to construct
the 3-D SUTRA finite element mesh and to generate the necessary input files containing node-
40
by-node aquifer properties, boundary conditions, initial conditions, and fluid source/sink terms
such as recharge and pumping (i.e. extraction). The procedure of constructing the model is as
follows:
a) Draw the domain of the island – this is accomplished by importing the island GIS
shapefile generated within ArcMap and then drawing the outline of the island. The effect
of the ocean is simulated in the model, with the model domain extending out to the ocean
on all sides of the island. The ocean domain is big enough to minimize the influence of
the boundary conditions on groundwater dynamics within the island aquifer. In this study,
the size of ocean is set as double the size of the island, with the shape similar to the
island.
b) Generate the finite element mesh –SUTRA uses irregular quadrilateral elements, which
are refined locally along the top of the model and along the coastal boundary where large
changes in groundwater hydraulic head and salt concentration are expected during the
model simulations. The elevation of island surface in the model is set to be 0, since the
study only simulates saturated flow (i.e. aquifer below mean sea level) is simulated in this
study. The slope on the coast is set to 1% which is the most common island slope in
Maldives. The surface mesh and vertical discretization of the mesh are shown in Figure
13. The layer thickness near surface is < 1m while layers toward the bottom of the mesh
are between 20 – 30 m.
c) Aquifer properties - the layers of the mesh are assigned to either the Holocene aquifer (0
m to 20 m below sea level) or the Pleistocene aquifer (20 m to 200 m below sea level).
The Holocene aquifer is assigned a porosity of 0.2, whereas the Pleistocene aquifer is
assigned a porosity of 0.3. Its permeability is much higher than the first layer. (Table 8)
41
Table 8: Properties of the atoll island aquifer system, including general properties, and properties for the Holocene
and Pleistocene aquifer units for 3-D model construction
Parameters
Value
Units
Source
General Aquifer Properties
Compressibility of Porous
Matrix
Specific Yield
Longitudinal Dispersivity αL
Transverse Dispersivity αT
1.00 x 10-9
0.20
6.0
0.05
m2/N
m3/m3
m
m
Peterson and Gingerich 1995
Griggs and Peterson 1993
Holocene Aquifer
Holocene Porosity
Holocene thickness
Holocene horizontal K
Holocene vertical K
0.2
13 to 18
75
15
m3/m3
m
m/d
m/d
Anthony 1997
Hamlin and Anthony 1987
Calibrated value (this study)
Calibrated value (this study)
Pleistocene Aquifer
Pleistocene Porosity
Pleistocene thickness
Pleistocene horizontal K
Pleistocene vertical K
0.3
485
5000
1000
m3/m3
m
m/d
m/d
Swartz 1962
Griggs and Peterson 1993
Oberdorfer et al. 1990
Oberdorfer et al. 1990
d) Boundary conditions and Initial Conditions – the mesh nodes within the ocean domain
have a specified pressure which represents the pressure from overlying seawater. The
nodes along the island top (i.e. the top layer of the mesh) are a fluid flux boundary and
represent locations of recharge to the freshwater lens. The freshwater inputs to each node
are the product of the recharge rate and the spatial area of the node.
Fifteen meshes were constructed to create a mesh that provided minimized error of the results
as well as the model run-time. The final mesh contained elements of 45 m within the interior of
the island and 18 m along the coast. It has thirty layers for the upper (Holocene) aquifer and
eight layers for the lower (Pleistocene) aquifer, with a total of 480,000 nodes. Part of the process
of making the mesh is showed in the following Figure 14. Using a time step of one day,
approximately 10 days are required to run a 39-year simulation on a 8-professor, 6G computer.
42
Figure 13. Top view and cross section view of the mesh of Gan. The island surface is the red area in Graph A
with elevation of 0 m. Other colors presents the elevation change of the islands extent under the ocean.
43
Figure 14. Graph A, B, C, D shows the process of making the mesh. In Graph A, the mesh is fine everywhere
and even more fine for island surface but it takes 2.3 minutes to run one time step. Graph B decreased the size of
the mesh but it still take about 1.7 minutes/ time step. In Graph C, it remains the fine grid size near the coast but
coarsen the grid far from coast. It takes about 1.3 minutes/ time step. In the last graph, it enlarged the ocean grids.
The farther the grid is from coast, the coarser they are. The grids in the center are coarsened but the ones around
coast line remain fine. It takes about 1minutes/ time step.
3.2.1.2 Pumping in Model and SUTRA code modification
In the island of Gan, groundwater recharges from rainfall and is pumped out for
anthropogenic use. There are three villages in Gan. One is in the northwest of the island, one is in
44
the mideast and the other is located in the south. The estimate of total number of occupied houses
is 691, as of October 2009, and each house has one or two pumping wells (Bangladesh, 2010).
That mean there are about 700-1300 pumping wells. Adding each individual well to the model is
too time consuming. So, rather than specify individual pumping nodes, this study defines
pumping areas (Figure15A). The nodes in pumping areas are all considered as pumping nodes
with uniform pumping rates. The three main pumping areas correspond to the three villages in
Gan. Three polygon objects are created to simulate pumping (Figure15B). Based on the
Bangladesh report, estimated groundwater usage is 120L/day/person. The total water pumped out
for each village is approximated as the product of pumping rate and population. As a result, the
pumping rate in each pumping area is total water usage divide by the number of nodes in
pumping area (See equations below).
�
�
�
�
�������
�
=
�
=
�
/ / ×
�
�
�
�
��
�������
�������
�������
The SUTRA executable provided within the downloaded software package from the USGS is
a 32-bit software that can only allocate a maximum of 2 GB instant memory for each model
simulation. When the mesh is too large and the running process required more than 2GB, it
cannot be executed. Therefore, GFortran was used to recompile SUTRA code into a 64-bit
executable. Also, subroutines within the code were implemented to allow for time-dependent
fluid source/sinks terms (i.e. recharge and pumping).
45
A
B
Figure 15. Pumping area (Green colors) in the model (Figure 15B) that represents the pumping in Gan (Figure
15A). Three green areas simulate the pumping in three villages in Gan.
3.2.3 Model Calibration
Even though all the values of aquifer properties are from calibrated 2-D SUTRA model, 3-D
model might be different. Geological investigation also only gave general range of these
properties. In consideration of many factors, vertical permeability is the key factors to determine
freshwater lens thickness—especially in the Holocene aquifer, which contains most of the
freshwater.
46
There is limited historical data to calibrate the model. The Bangladesh Company did a
groundwater survey in Gan in December 2009, which gave freshwater lens thickness in 25
locations (Table 1) and estimated freshwater volume. The model uses historical rainfall data from
1998 to 2012 as an input. Recharge is calculated using the same method as Chapter II. To
calculate the pumping rate (See Section 2.1.1), the population figures in 2006 are used. Thus, the
pumping rate in the village Thundi, Mathimaradhoo and Mukurimagu are 7.80E-03 kg/s, 1.54E03 kg/s and 4.19E-03 kg/s, respectively. The simulation starts with the model completely filled
with seawater, with a freshwater influx over many years, and runs until the freshwater lens is
stable. Then, the results of that stable freshwater lens are used as the initial condition for the final
model. The final run uses calculated recharge from 1998-2012 and transient pumping rate. The
results of freshwater lens thickness taken in December, 2009 are compared to the investigated
results at the same location. The simulation attempts many vertical permeability ranges from 110 m/day with an interval of 1 m/day. The value which best matches the practical data is
considered the calibrated model.
3.2.4 Estimating the Effect of Future Climate Scenarios on Gan’s Freshwater Lens
This section lists the methods which estimate the effects of rainfall pattern, population growth
and sea-level rise on the groundwater resources in Gan. All the GCMs used for each of the
simulations are listed in Table 9.
47
Table 9: All the GCMs used for simulations of each scenario.
Rainfall Pattern
Conservative Pumping
Aggressive Pumping
Sea-level Rise
Sea-level Rise and Aggressive
Pumping
Rcp2.6
CESM1-CAM5
MIROC5
IPSL-CM5A-LR
CCSM4
GFDL-ESM2G
CESM1-CAM5
MIROC5
IPSL-CM5A-LR
CCSM4
GFDL-ESM2G
CESM1-CAM5
MIROC5
IPSL-CM5A-LR
CCSM4
GFDL-ESM2G
Rcp8.5
CESM1-CAM5
MIROC5
CCSM4
IPSL-CM5A-MR
GISS-E2-R p1
GFDL-ESM2G
GFDL-CM3
CESM1-CAM5
MIROC5
CCSM4
IPSL-CM5A-MR
GISS-E2-R p1
GFDL-ESM2G
GFDL-CM3
Steady recharge
CCSM4
IPSL-CM5A-LR
3.2.4.1 Future Estimated Precipitation
GCMs are used to estimate future climate change. In Chapter II, five better-performing
GCMs are selected for in Region 2— where Gan is— for scenario Rcp2.6 and seven GCMs are
selected for scenario Rcp8.5. The same procedures are used to downscale to daily rainfall data
and calculated recharge (Figure 16A) using mass balance. Time series of predicted rainfall data
for 2012-2050 from these five GCMs are plotted in Figure 16A. SUTRA uses this data to predict
future lens thickness. The output gives salt concentrations at all locations for every two months.
The freshwater lens is limited at 0.00089 kgsalt/kgwater. The average lens thickness of whole
islands is calculated by MATLAB.
48
A
30
Monthly Rainfall for five GCMs for Rcp2.6
Rainfall (mm/day)
25
20
15
10
5
0
2013
2018
CESM1-CAM5
Rainfall (mm/day)
B 20
2023
2028
2033
MIROC5
IPSL-CM5A-LR
2038
2043
CCSM4
2048
GFDL-ESM2G
Monthly recharge for five GCMs
15
10
5
0
2013
2018
CESM1-CAM5
2023
2028
2033
MIROC5
IPSL-CM5A-LR
2038
2043
2048
CCSM4
GFDL-ESM2G
Figure 16. Time series plots of precipitation (A) and corresponding recharge (B) from five selected GCMs with
scenario Rcp2.6 from the year of 2013 to 2050 for island of Gan.
3.2.4.2 Future Estimated Pumping
Pumping needs varies with population. The Maldives has a high annual population growth
rate of 1.76%. The island of Gan is even worse. The average annual growth rate from 2006 to
2010 was 13.5%. The Government of Maldives (GoM) estimates that the population could rise to
20,000 (Bangladesh, 2010). Gan is the largest island in Maldives. Its population density (11.2 /ha)
49
is much lower than the capital Malé (260/ha). As population stress increasing in the future, it is
very possible that more people will move to Gan. In this study, in order to estimate future
pumping, conservative and aggressive estimation are both considered for population growth. The
conservative estimate uses a population growth of 1.76%; the aggressive estimation uses 9%.
Population growth and pumping rates are plotted in Figure 17. These two different pumping rates
both run with all GCMs from scenario Rcp2.6 and Rcp8.5. the resulting data is compared with
regards to the effects on groundwater resources. Also, one example that uses climate data from
CCSM4 is used to compare no pumping with conservative pumping and examine the pumping
effects on groundwater resources.
Figure 17. Estimated conservative and aggressive population growth and pumping rate in the future for Gan
3.2.4.3 Future Sea-Level Rise
Sea-level rise has been a big issue for Maldives. As the sea level rises, massive land areas in
Maldives are inundated by seawater, especially for Maldives. It has the lowest average elevation
of about 1.5 m above sea level and the slope of the islands is about 1:100. The sea-level is
predicted to rise a half meter for Maldives during 21st century (Woodworth, 2005). As a result,
about 50 m of shoreline would be inundated. Some islands may disappear.
For Gan, the sea level rise is a serious problem for its fresh groundwater. As the sea level
50
rises, the seawater pressure underground increases and pushes up the freshwater lens upward.
The island’s shrunken surface would discharge more freshwater to the sea and would receive less
precipitation. The model simulates the effect of sea level rise on groundwater by building a new
model with a smaller islands surface (Figure 18). About 50m of coastline is lost to the ocean in
the model. The green line is the surface before sea-level rise, whereas the red one marks the
surface after. The sea level rise decrease land area by about 12.5%. In that simulation, pumping
is not applied and rainfall rate is kept steady in order to isolate the effects of sea level rise. The
rainfall rate is about 2.73 mm/day which is the average of the rainfall rate from 2011 to 2050.
Two models are simulated to contrast with each other: before and after sea level rise. The models
assume the island is full of seawater as initial condition, then run until a steady state for about 20
years. The results give us the percent of freshwater lens volume lost according to the amount of
land area.
3.3 Results
3.3.1 3-D views of simulation results
The simulation results give the salt distribution at each mesh node in each time step. The
Figure showed below is a 3-D view of the salt distribution in the groundwater of island Gan
(Figure 19). In Figure 19A, it shows the whole island model grid. The red part is seawater, which
has the highest salt concentration. The blue part is mostly fresh water, which has a very low salt
concentration. Other colors, like yellow and green, represent the parts with salt concentration
between seawater and freshwater. A mix of seawater and freshwater is not potable. Figure B and
C show the cross section of the island. The blue part is the freshwater lens and its thickness. It
gives the lens thickness in each point of the island. Freshwater volume can be calculated from
that under certain limits of salt concentration for each time step.
51
Figure 18. Land surface comparison before and after sea-level rise. The blue line is the contour before sea-level rise
whereas the red one is after sea-level rise.
52
Figure 19. Three dimensional review of the SUTRA modeling results. The color represents the salt
concentration. The red color has highest salt concentration that represents seawater. The blue color is freshwater.
Other colors are the mixture of fresh and seawater.
3.3.2 Calibration Results
The model is calibrated with different vertical permeability in the time of December 2009.
Few plots of observed freshwater lens thickness on certain points of Gan versus modeled values
are showed as below.
�
=
/ �� looks the best match for historical data (Figure 20). With
the best match, the difference of the mean of the observed and modeled value is pretty small
about 0.033 m, which can be reasonable approximation to be unbiased, but the relative error is
still large with about 58%. The total estimated freshwater is very close to field measurement
estimate, which is about 14300 million liter versus 14200 million m3 from groundwater
investigation report (Bangladesh Ltd., 2010). Overall, the model might perform better if there are
more available historical measurements.
53
Figure 20. Observed vs. modeled lens thickness plot with different vertical permeability. The red line the y=x
line. The �� = � �/� shows the best match.
3.3.3 Pumping effects
3.3.3.1 Pumping vs. Non pumping
One model is simulated to generally show the pumping effect on freshwater volume by using
one of the GCMs (CCSM4). Figure 21 shows the freshwater volume time series plot and
compares pumping to no pumping. With the conservative pumping, volume decline is observed
at the beginning of pumping and increased later as pumping increased. Comparing to nonpumping simulation, the average loss of fresh groundwater volume is about 6.77% and the
highest is 10.5% (Figure 22). At the end of 2050, the total loss of fresh groundwater from
pumping is 1.73 million m3. However, the total water pumped out for 39 years during 2012-2050
are much greater than that, which is 10536 million m3. The water loss of fresh groundwater from
54
pumping is less than 0.02% of the water pumped out, which is not significant at all. In addition,
in the beginning five or six years, the losses increase linearly. However, after that, the losses are
not a linear progression because it changes with wet and dry seasons.
Pumping effects on Freshwater Volume
20
Million m3
18
16
without pumping
14
with pumping
12
10
2012
2019
2026
2033
2040
2047
Figure 21. Comparison of freshwater volume time series plots from the scenarios with pumping and
without pumping.
Volume Losses from Pumping
Million m3
2
1.6
1.2
0.8
0.4
0
2012
2019
2026
2033
2040
2047
Figure 22. Time series plot of freshwater volume plot from pumping. The loss is the freshwater volume
difference between pumping scenario and non-pumping scenario.
3.3.3.2 Conservative pumping
55
The conservative pumping demands grow at a rate of 1.76%, due to population growth. The
results show that the freshwater volume increases in the future even with more pumping. The
Figure 23A is the time series plot of freshwater lens volume for five different GCM scenarios.
The GCM scenarios give similar results. CCSM4 and MIRCO5 are close to each other, but there
is a large difference between CESM-CAM5 and IPSL-CM5A-LR. The range of the volume is
within about 10 to 20 million m3 and the average is about 15.3 million m3. At the end of 2050,
three of them reached a point of 16 million m3 and the other two meet at 15 million m3.
Fluctuation of volumes corresponding to wet and dry season are observed. CCSM4, MIRCO5
have small fluctuations whereas the other three have much larger fluctuations. For example,
GFDL-ESM2G has 3-6 large fluctuations that some of them are over a period of ten years. This
causes a larger range for freshwater volume and more extreme wet and dry season,
correspondingly. A slight uptrend in freshwater volume is observed for all the GCM scenarios.
These increases of freshwater volume, for GCMs are 6.9%, 5.6%, 15.9%, 15.1% and 18.1%,
respectively, with an average increase of 12%. The Figure 23B shows the average lens thickness
of the whole island for the simulation time period, 2012-2050. The range of average lens is from
5 meter to 8 m and the average of all scenarios across the simulation period is about 6.6 meters.
The fluctuation of lens thickness can be as larger as 2 meters. Overall, with conservative
pumping, the fresh groundwater would still satisfy the pumping needs in the future.
56
Freshwater Volume (Conservative pupming)
A
25
Million m3
20
CCSM4
15
CESM-CAM5
10
GFDL-ESM2G
IPSL-CM5A-LR
5
MIRCO5
0
2012
2019
2026
2033
2040
2047
Freshwater Len Thickness (Conservative Pumping)
B
9.0
Thickness (m)
8.0
CCSM4
7.0
CESM-CAM5
6.0
GFDL-ESM2G
5.0
IPSL-CM5A-LR
4.0
MIRCO5
3.0
2012
2019
2026
2033
2040
2047
Figure 23. Time series plots of freshwater volume and lens thickness for all selected GCMs
3.3.3.3 Aggressive pumping
With aggressive pumping, the pumping rates increase by 9% per year. A large decrease in
fresh groundwater volume is predicted (Figure 24). All GCMs have decreasing trend. By the end
of 2050, the freshwater volume would decrease by 23.7% on average for scenario Rcp2.6. The
average freshwater value is about 10.6 million m3. The average lens drops from 6 m to 5.56
meters. However, the lens area, which held freshwater, shrank about 20% from 6
4.79
to
, parts of shoreline area do not have fresh groundwater anymore. The results for
57
scenario Rcp8.5 are more diverse. The time series pattern of freshwater volume for all GCMs is
less consonant than Rcp2.6. and the range for volume at the end of 2050 is much larger. The
average decrease of freshwater volume is about 20.88%. The lens area shrinks about 21.6% at
the end of simulation. Overall, with aggressive pumping, the freshwater lens would essentially be
destroyed.
Freshwater Volume (Rcp2.6)
Million m3
A
20
18
16
14
12
10
8
6
4
2012
CESM-CAM5
GFDL-ESM2G
IPSL-CM5A-LR
MIRCO5
2019
2026
2033
2040
2047
Freshwater Volume (Rcp.8.5)
B
Million m3
CCSM4
20
18
16
14
12
10
8
6
4
2012
CCSM4
CESM-CAM5
GFDL-CM3
GFDL-ESM2G
GISS-E2-R-p1
IPSL-CM5A-MR
2019
2026
2033
2040
58
2047
C
Freshwater Lens Thickness (Rcp 2.6)
Thickness (m)
8.0
7.0
CCSM4
6.0
CESM-CAM5
5.0
GFDL-ESM2G
IPSL-CM5A-LR
4.0
MIRCO5
3.0
2012
2019
2026
2033
2040
2047
Figure 24. Plot A and B show the time series plots of freshwater volume from scenario Rcp2.6 and Rcp8.5. Plot
C shows the time series plot of freshwater lens thickness.
3.3.4 Sea-level rise effects
The land area of Gan was about 6.57
in 1998, but it will drop to 5.75
during the
21st century. The freshwater volume will decrease in response to smaller land area. With an
inundation of 12.5% of Gan’s land area, it loses 13.73% of freshwater volume. Figure 25 shows a
time series plot which compares freshwater volume before and after sea level rise. At the
beginning of simulation with 100% seawater, there is no fresh water. As rainfall is applied on the
models, the freshwater volume starts to increase. For the first few years, two models have a
similar increasing trend. After seven years, a gap develops between them. After about fourteen
years, the trends approach a steady gap. At the end of simulation, the before sea level rise model
has a freshwater volume of 16.1 million m3 whereas after sea rise model has 13.9 million m3. To
conclude, the effects of sea-level rise on freshwater lens in Gan are bigger than moderate
population growth but smaller than aggressive population growth.
59
Million m3
Comparison of Freshawter Volume
18
16
14
12
10
8
6
4
2
0
Before sea-level rise
After sea-level rise
1
4
7
10
13
Simulation Years
16
19
Figure 25. Time series plot comparison of freshwater volume between simulations before sea-level rise and
after sea-level rise.
3.3.5 Combined effects of sea-level rise and aggressive pumping
Under the impact of sea level rise and aggressive pumping, the freshwater volume drops
tremendously. At the end of 2050, the freshwater volume decreased about 87% from about 14.1
to 1.8 million m3. The area that has freshwater is decreased to 10% of the total land area. The
average lens thickness decreased from 6.8 m to about 1.1 m. With less than 2 meters of lens
thickness, pumping is not viable. The freshwater lens is effectively destroyed even with some
remaining fresh water. These results are only valid at the year of 2050.
3.4 Discussion
3.4.1 Advantage and disadvantage of the model
The model mesh is fine enough to give a good resolution for the results but model runs
slowly. The model mesh is built in ModelMuse. It has approximately a half-million elements or
nodes with thirty layers in the Holocene aquifer and eight layers in Pleistocene aquifer. It takes a
60
long time to run and lots of computation memory. For the simulation from 2012-2050 with about
14235 daily time steps, it takes about two weeks to run and 1.7 GB computer memory. This
study ran about forty models. Lots of computers resources or super computers are needed to do
this research. Another way to shorten the running time would be to improve the SUTRA model
performance such as optimizing the algorithm or changing it to parallel processing, which may
be too difficult. However, improvement of model performance is necessary and can be done in
future modeling.
The boundary condition is simplified to ease processing. The rainfall rate is assumed to be
homogenous for the island. The model top is mean sea level, not the island’s surface, which
means unsaturated zones are not simulated. The island’s topography is unknown, so the slope of
shoreline is estimated. Technically, the north and south ends are connected with other islands.
However, they are assumed surrounded by seawater. This can affect the model results. All these
simplifications of boundary conditions can be improved in the future modeling,
The assumption of homogeneity of the two aquifers would also affect the model results. And
it can be significant. With limited geographic data, it is impossible for us to accurately estimate
the aquifer properties. Accordingly, more geographic investigation of freshwater lens is
necessary to improve the model performance. The model calibration is limited by the available
data. Only one groundwater investigation in 2009 has estimated the freshwater lens thickness.
Obviously, more field data are needed to improve calibration results too. Current field data could
be used to validate the model. Overall, more investigation and field data are needed in the future.
61
3.4.2 Effects of changing rainfall pattern
The fresh groundwater would increase in the future according to model results. This is
because the rainfall is predicted to increase in the future by GCMs. As a result, there will actually
be more fresh groundwater available for the island of Gan. This conclusion also coincides with
2-D modeling. However, more rainfall could contribute to more frequent extreme events.
Extreme rainfall or dry seasons are predicted to be likely in the future too.
3.4.3 Effects of pumping
3.4.3.1 Pumping vs. non pumping
A mild increase of pumping rates does decrease the freshwater volume. However, compared
to the total fresh water volume pumped out, the decrease freshwater volume is insignificant. The
main reason is that the water discharge to the sea is being replaced by the pumping. In other
words, much less water is discharged to the sea at the edge of the island when it is pumping
because the water table stays low. Otherwise, the freshwater from rain could discharge to the sea
without being used. Hence, it is safe to pump a reasonable amount of fresh groundwater for use
by residents. In addition, it is interesting to see the losses (Figure 22) change seasonally instead
of by a positive linear progression.
3.4.3.2 Conservative pumping vs. aggressive pumping
With conservative pumping, the freshwater lens keeps its shape and is not affected much by
the pumping. The population growth assumed in this scenario is 1.76% per year. At the end of
2050, the population doubles, which is not a small amount of stress. However, the population
stress does not outweigh the projected increase in rainfall: the freshwater volume increases even
with conservative pumping stress. The average lens thickness remains around 6-7 m, which is
thick enough for pumping. This indicates that with high population growth rate of 1.76% in Gan,
62
there is enough fresh groundwater for their usage for next forty years. However, aggressive
pumping with a population growth rate of 9% cause severe seawater intrusion: the freshwater
would become exhausted. Moreover, the land area near the shoreline would run out of fresh
groundwater first. All three villages are very close to the shoreline, so pumping issues would
occur very quickly. This scenario requires a shallower pump or a central location for pumping on
the island. Even though the freshwater volume decreases a lot, the lens is thick enough for
pumping. The population growth did stress on the groundwater in Gan. The assumed population
growth in Bangladesh Consultants reports may be too high, but it could happen in the future.
Consider the capital Male: it has nearly 200,000 people and they run out of fresh groundwater.
To avoid this situation happens again, sever results from population stress needs to be taken into
account by the government of Gan. The number of people who live in Gan should be limited for
the sake of groundwater quantity and quality.
3.4.3.3 Sea-level rise
Sea-level rise does have a huge effect on the groundwater in Gan. First, the shoreline of Gan
will be pushed inward 50-80 m by seawater. All three villages are very close to the shoreline and
the closest roads or houses are within 50 m of shoreline. If sea-level rise happens, these houses
would have to be moved. Second, with 12.5% of land area inundated, 12.5% less rainfall
recharged is received by the smaller land area. Third, freshwater is lost because the freshwater is
pressurized by the raised sea level. The advantage of the model is that it can quantify the
freshwater losses from sea-level rise. However, the model simplified the sea-level rise process.
In reality, the sea level rises slowly and continuously. So, the land area changes alongside the sea
level change. Nevertheless, in this study, the model only simulates the situation at the end of 21st
century by making a smaller land surface model while the rainfall is assumed steady. The results
63
can only give the final stage of sea level rise effects. The process of sea level rise is not
simulated, which could have impacts on the final results.
The combined effects of sea-level rise and aggressive pumping will exhaust freshwater
resources in Gan very quickly. In a short time there would be no potable water for drinking or
domestic use. The government of Gan should be prepared to preserve groundwater resources to
prevent this scenario from happening.
64
CHAPTER 4: SUMMARY
This study helps assess current and future groundwater resources in Maldives by using 2-D
and 3-D numerical modeling. 2-D modeling gives an overall assessment of fresh groundwater
quantity for 4 different sizes of islands (200, 400, 600 and 1100 m) in Maldives. Estimations of
lens thickness and its fluctuation from the year of 2012 to 2050 are provided by 2-D modeling. A
mild increase in lens thickness in the future is found under the effects of changing rainfall
pattern. 3-D modeling is used to evaluate the effects of changing rainfall pattern, increasing
pumping and sea-level rising on the groundwater resources for a specific island, Gan. Same
results from 2-D modeling are found on the effects of changing rainfall pattern in 3D modeling.
An increase in fresh groundwater volume is estimated in the future for Gan since more rainfall is
predicted by GCMs. In terms of pumping, conservative pumping with a normal population
growth of 1.76%/yr would not have negative effects on the quantity of groundwater resources.
Instead, freshwater groundwater can be a very valuable water resource for Gan with reasonable
pumping. More freshwater could be pumped out for usage instead of discharging to the sea
without destroying the freshwater lens. However, with aggressive pumping, it will exceed the
capacity and sustainability of the groundwater. The lens would essentially be destroyed
essentially. Under the effects of sea-level rise, the groundwater volume shrinks as land area
shrinks. It will have a significant effect on groundwater resource in atoll islands if it happens.
The worst case is that sea-level rise and aggressive pumping all happens. In this case, the
groundwater resources in Gan are exhausted within few years. Overall, this study gives an
estimation and evaluation of major impacts on groundwater resources for the future of Maldives.
65
This study would also help the water management for the government of Maldives. The
overall assessment and estimation of groundwater resources can give them a general
understanding of freshwater quantity in different islands. Especially, estimation of future
groundwater resources can help with future water resources planning. Residents also can plan
the pumping according to the seasonal fluctuation of lens thickness in case of pumping seawater
out. The results from 3-D modeling give them a more clear and mathematic understanding of
effects of pumping and sea-level rising on groundwater resources. With these quantified impacts,
it makes easier for the government of Gan in Maldives to manage pumping, population and water
resources. Knowing the negative results of aggressive pumping and sea-level rise, they can come
up with better solutions of water management to prepare for that happening. Modeling results
can develop improved the understanding of groundwater resources in Maldives and also help
with water supply management for the Ministry of Environment, Energy and Water (MEEW) and
Male Water and Sewerage Company (MWSC).
The 3-D modeling can be applied for other islands in the Maldives or other atolls. 3-D
modeling is better simulates on radial pumping than 2-D modeling. However, there are still lots
of things to be improved. The model can be improved significantly by inputting more detail
information about geology, topography and local rainfall data of the island. Unsaturated zone can
also be simulated in the future. Evapotranspiration can be calculated more accurately by
knowing the plants or vegetable type on the islands. The tide can be simulated by applying
transient pressure boundary along the shoreline. The simulation methods can be improved by
specifying pumping location and pumping rates in different time, changing the sea-level rising to
a transient process. The future research would focus on the accuracy of the modeling.
66
REFERENCE
Ablain, M., Cazenave, A., Valladeau, G., & Guinehut, S. (2009). A new assessment of the error
budget of global mean sea level rate estimated by satellite altimetry over 1993–2008.
Ocean Science, 5(2), 193-201.
AFD (2007). Sewerage System Creation in 4 Islands, Maldives Tsunami Infrastructure
Rehabilitation Project, Final Report. Prepared by BRL Ingénierie for Agence Française
de Développement, November 2007.
Ayers, J.F., and H.L. Vacher (1986), Hydrogeology of an atoll island: A conceptual model from
detailed study of a Micronesian example. Ground Water 24, 2-15.
Bailey, R., Khalil, A., & Chatikavanij, V. (2014a). Estimating transient freshwater lens dynamics
for atoll islands of the Maldives. Journal of Hydrology, 515, 247-256.
Bailey, R., Khalil, A., & Chatikavanij, V. (2014b). Estimating Current and Future Groundwater
Resources of the Maldives. JAWRA Journal of the American Water Resources
Association, 51(1), 112-122.
Bailey, R.T., Jenson, J.W., and A.E. Olsen (2009), Numerical Modeling of Atoll Island
Hydrogeology. Ground Water 47, 184-196.
Bangladesh Consultants, Ltd. 2010a. Groundwater Investigations Report for GDh. Thinadhoo,
Ministry of Housing, Transport and Environment Republic of Maldives.
Bangladesh Consultants, Ltd. 2010b. Groundwater Investigations Report for L.Gan, Ministry of
Housing, Transport and Environment Republic of Maldives.
Bangladesh Consultants, Ltd. 2010c. Groundwater Investigations Report for N.Holhdhoo,
Ministry of Housing, Transport and Environment Republic of Maldives.
Bangladesh Consultants, Ltd. 2010d. Groundwater Investigations Report for N.Velidhoo,
Ministry of Housing, Transport and Environment Republic of Maldives.
Barthiban, S., Lloyd, B., & Maier, M. (2012). Sanitary Hazards and Microbial Quality of Open
Dug Wells in the Maldives Islands. JWARP Journal of Water Resource and Protection,
4(7), 474-486.
67
Beswick R. (2000). Water Supply and Sanitation, A Strategy and Plan for the Republic of
Maldives, Parts 1 & 2, Ministry of Health, Republic of Maldives, March 2000.
Brown, B. E. & Dunne, R. P. (1988). The environmental impact of coral mining on coral reefs in
the Maldives. Environ. Conserv., 15, 159-66.
Church, J.A., N. White and J. Hunter (2006), Sea level rise at tropical Pacific and In-dian Ocean
islands. Global Planet.Change , 53 , 155-168.
Coastal Aquifers, U.S. Geol. Surv. Water Supply Pap. 1613-C, 35-69, 1964.
Coe, R., and R. D. Stern. "Fitting models to daily rainfall data." Journal of Applied Meteorology
21.7 (1982): 1024-1031.
Comte, J. C., Join, J. L., Banton, O., & Nicolini, E. (2014). Modelling the response of fresh
groundwater to climate and vegetation changes in coral islands. Hydrogeology Journal,
22(8), 1905-1920.
Cox PM, Betts RA, Jones CD et al. (2000) Acceleration of global warming due to carbon-cycle
feedbacks in a coupled climate model. Nature, 408 , 184-187
Dickinson, W.R. (2004), Impacts of eustasy and hydro-eustasy on the evolution and landforms of
Pacific atolls. Palaeoecology 213, 251-269.
Drabbe, J., and Badon Ghyben, W., 1889, Nota in verband met de voorgenomen putboring nabij
Amsterdam [Notes on the probable results of the proposed well drilling near Amsterdam]:
The Hague, Koninkl. Inst. Ing. Tijdschr., 1888-89, p. 8-22.
Falkland, A., & Custodio, E. (1991). Conditions for freshwater occurrence in small islands. In
Hydrology and water resources of small islands: A practical guide: A contribution to the
International Hydrological Programme, IHP-III, Project 4.6. Paris: Unesco.
Falkland, A.C. (1994), Climate, Hydrology and Water Resources of the Cocos (Keeling) Islands.
Atoll Research Bulletin No. 400, Smithsonian Institute, Washington, D.C., 23 pp.
Falkland, A.C. 1983. Christmas Island (Kiritimati) Water Resources Study. Vol. 1. Victoria,
Australia: Australian Department of Housing and Construction.
68
Falkland, T. 2000. Report on Groundwater Investigations in Southern Development Region
(ADB Regional Development Project). Report for Ministry of Planning and National
Development.
Falkland, T. 2001. Report on Groundwater Investigations in Northern Development Region
(ADB Regional Development Project). Report for Ministry of Planning and National
Development.
Fu, G., Liu, Z., Charles, S., Xu, Z., & Yao, Z. (2013). A score-based method for assessing the
performance of GCMs: A case study of southeastern Australia. Journal of Geophysical
Research: Atmospheres J. Geophys. Res. Atmos., 118, 4154-4167.
Ghassemi, F., Alam, K., & Howard, K. (2000). Fresh-water lenses and practical limitations of
their three-dimensional simulation. Hydrogeology Journal, 8(5), 521-537.
Ghassemi, F., Jakeman, A. J., Jacobson, G., & Howard, K. W. F. (1996). Simulation of seawater
intrusion with 2D and 3D models: Nauru Island case study. Hydrogeology Journal, 4(3),
4-22.
Ghassemi, F., Molson, J. W., Falkland, A., & Alam, K. (1998). Three-dimensional simulation of
the Home Island freshwater lens: preliminary results. Environmental modelling &
software, 14(2), 181-190.
Gingerich, S. B., & Voss, C. I. (2005). Three-dimensional variable-density flow simulation of a
coastal aquifer in southern Oahu, Hawaii, USA. Hydrogeology Journal, 13(2), 436-450.
Glover, Robert E., The pattern of fresh-water flow in a coastal aquifer. Sea Water in Coastal
Aquifers, U.S. Geol. Surv. Water Supply Pap. 1613-C, 35-69, 1964.
GoM (2010). Website “Isles” (http://isles.egov.mv), Government of Maldives.
Griggs, J.E. and Peterson, F.L. (1993), Ground-water flow dynamics and development strategies
at the atoll scale. Ground Water 31(2), 209-220.
GWP Consultants (2006), Maldives water and sanitation authority five year activity plan 20062010.
Hamlin, S.N. and S.S. Anthony (1987), Ground-water resources of the Laura area, Majuro Atoll,
Marshall Islands. USGS Water Resources Investigation Report 87-4047.
69
Han, W., Meehl, G.A., Rajagopalan, B., Fasullo, J.T., Hu, A., Lin, J., Large, W.G., Wang, J.-w.,
Quan, X.-W., Trenary, L.L., Wallcraft, A., Shinoda, T., and S. Yeager (2010), Patterns of
Indian Ocean sea-level change in a warming climate. Nature Geoscience Letters, doi:
10.1038/NGEO901.
Hay, J. (2006). Climate Risk Profile for the Maldives. Maldives: National Disaster Management
Center.
Herman, M.E., and S.W. Wheatcraft. 1984. Groundwater dynamics investigation of Enjebi
Island, Enewetak Atoll: An interpretive computer model simulation. In Finite Elements in
Water Resources: Proceedings, vol. 5, 133–142, ed. J.P.Laible, C.A.Brebbia, W. Gray, G.
Pinder. London, England: Springer-Verlag.
Herzberg, Alexander, 1901, Die Wasserversorgung einiger Nordseebader [The water supply on
parts of the North Sea coast]: Munich, Jour. Gasbeleucht. u. Wasserversorg., v. 44, p.
815-819, 842-844.
Hogan, P.J. 1988. Modeling of fresh water-saltwater interaction on Enjebi Island, Enewetak
Atoll. M.S. thesis, Department of Geology, San Jose State University, San Jose,
California.
Hunt, C.D., Jr. 1997. Hydrogeology of Diego Garcia. In Geology and Hydrogeology of
Carbonate Islands. Developments in Sedimentology 54, eds. H.L. Vacher and T. Quinn,
909-931.
J C Pernetta and D.L. Elder, 1990, Climate, Sea Level Rise and the Coastal Zone: Management
and Planning for Global Changes, International Union for the Conservation of Nature
and Natural Resources, Gland.
Karthikheyan, T. (2010). Environmental Challenges for Maldives. South Asian Survey, 343-351.
Kench, P.S., McLean, R.F., Brander, R.W., Nichol, S.L., Smithers, S.G., Ford, M.R., Parnell,
K.E., and M. Aslam (2006), Geological effects of tsunami on mid-ocean atoll islands:
The Maldives before and after the Sumatran tsunami. Geology, 34(3), 177-180.
Ketabchi H., Mahmoodzadeh D., Ataie-Ashtiani B., Werner A. D., and Simmons C. T. (2014),
Sea-level rise impact on fresh groundwater lenses in two-layer small islands, Hydrol.
Process., 28; pages 5938–5953, doi: 10.1002/hyp.10059
70
Kipp, K. L. (1987). HST3D; a computer code for simulation of heat and solute transport in threedimensional ground-water flow systems.
Lam, R.K. 1974. Atoll permeability calculated from tidal diffusion. Journal of Geophysical
Research 79, 3073–3081.
Lloyd, J.W., J.C. Miles, G.R. Chessman, and S.F. Bugg. 1980. A ground water resources study
of a Pacific Ocean atoll—Tarawa, Gilbert Islands.Water Resources Bulletin16, no. 4:
646–653.
Mather, J.D., 1975, Development of the groundwater resources of small limestone islands:
Quarterly Journal of Engineering Geology, v. 8, p. 141-150.
MEE, 2004. State of the Environment Maldives 2011. Ministry of Environment and Energy, The
Republic of Maldives, Male.
MEE, 2011. State of the Environment Maldives 2011. Ministry of Environment and Energy, The
Republic of Maldives, Male.
Meehl, G., Goddard, L., Murphy, J., Stouffer, R. J., Boer, G., Danabasoglu, G., . . . Stockdale, T.
(2009). Decadal Prediction: Can It Be Skillful? American Meteorological Society, 14671485.
Meurant, G. (1987). Multitasking the conjugate gradient method on the CRAY X-MP/48. Parallel
Computing, 5(3), 267-280.
Molson, J. W., & Frind, E. O. (1994). SALTFLOW: Density dependent flow and mass transport
model in three dimensions. User Guide, Waterloo Centre for Groundwater Research,
University of Waterloo, Ontario.
Mörner , N.-A., Tooley, M., Possnert, G., 2004. New perspectives for the future of the Maldives.
Global Planet. Change 40, 177–182.
MPHRE (1998). Fifth National Development Plan 1997-2000, Volume II, Ministry of Planning,
Human Resources and Environment, Government of Maldives. UNEP,(2005) PostTsunami Environmental Assessment, Male, Maldives
MPND, 2006, Census 2006, Ministry of Planning and National Development, Male, Republic of
Maldives.
71
Oberdorfer, J.A., P.J. Hogan, and R.W. Buddemeier. 1990. Atoll island hydrogeology: Flow and
freshwater occurrence in a tidally dominated system. Journal of Hydrology 120, 327–340.
Peinhardt, K. A. (2014). Climate Change Vulnerabilities: Case Studies of the Maldives and
Kenya. Honors Scholar Theses. Paper 383.
Pernetta, J.C. 1992. Impacts of climate change and sea-level rise on small island states. National
and international responses. Global Environmental Change 2(1), 19-31.
Presley, T. K. (2005). Effects of the 1998 drought on the freshwater lens in the Laura area,
Majuro Atoll, Republic of the Marshall Islands (No. 2005-5098). Geological Survey (US).
Rajasuriya AH, Zahir EV, Muely BR, Subramanian K, Venkataraman MVM, Wafar SM,
Munjurul Hannan Khan, Whittingham E (2000) Status of coral reefs in South Asia:
Bangladesh, India, Maldives and Sri Lanka. In: Wilkinson C (ed) Status of coral reefs of
the world: 2000. Aust Inst Mar Sci, Townsville, pp 95–116, http://www.reefbase.org/
references/ref_Literature.asp?searchactive=yes&ID=11755
Richmond, B.M., Jaffe, B.E., Gelfenbaum, G., Morton, R.A., 2006. Geologic impacts of the
2004 Indian Ocean Tsunami on Indonesia, Sri Lanka, and the Maldives. Zeitschrift für
Geomorphologie N.F., Suppl.-Vol. 146, 235–251.
Sovacool, B. K. (2012). Perceptions of climate change risks and resilient island planning in the
Maldives. Mitigation and Adaptation Strategies for Global Change, 17(7), 731-752.
Spennemann, D. (2006). Freshwater lens, settlement patterns, resource use and connectivity in
the Marshall Islands. Transforming Cultures eJournal, 1(2).
Srikanthan, R. "Stochastic generation of daily rainfall data." MODSIM2005, Melbourne, 12
(2005): 1915-1921.
Srikanthan, R., & McMahon, T. A. (2001). Stochastic generation of annual, monthly and daily
climate data: A review. Hydrology and Earth System Sciences, 653-670.
Stoer, J., & Bulirsch, R. (1980). Introduction to numerical analysis: New York, Springer-Verlag,
609p.
Taylor, K. E., Stouffer, R. J., & Meehl, G. A. (2012). An Overview of CMIP5 and the
experiment design. Bulletin of the American Meteorological Society 93, 485-498.
72
Terry, J. P., & Falkland, A. C. (2010). Responses of atoll freshwater lenses to storm-surge
overwash in the Northern Cook Islands. Hydrogeology Journal, 18(3), 749-759.
Terry, J. P., & Thaman, R. R. (2008). Physical geography of Majuro and the Marshall Islands.
The Marshall Islands: environment, history and society in the atolls. Faculty of Islands
and Oceans, The University of the South Pacific, Suva, Fiji, 1-22.
Terry, James P., and Ting Fong May Chui. "Evaluating the fate of freshwater lenses on atoll
islands after eustatic sea-level rise and cyclone-driven inundation: a modelling approach."
Global and Planetary Change 88 (2012): 76-84.
The World Bank. (2011). Maldives. Retrieved 11 10, 2012, from The World Bank:
http://www.worldbank.org/en/country/maldives.
Thurber, D., Broecker, W., Blanchard, R., & Potratz, H. (1965). Uranium-Series Ages of Pacific
Atoll Coral. Science 149, 55-58.
Todorovic, P., & Woolhiser, D. A. (1975). A Stochastic Model of n-Day Precipitation. Journal of
Applied Meteorology, 17-24.
Tsyban, A.V., J.T. Everett, and J.G. Titus (1990), World oceans and coastal zones. In: UNESCO
(1991), Hydrology and Water Resources of Small Islands: a practical guide. Studies and
Reports in Hydrology. No. 49. Falkland A. (editor), Custodio E. (1991) with
contributions from other authors. Unesco, Paris, France.
Underwood, M.R., F.L. Peterson, and C.I. Voss. 1992. Groundwater lens dynamics of atoll
islands. Water Resources Research 28, no. 11:2889–2902.
Urish, D. W. (1974), Fresh water on the coral atoll island. The Military Engineer, Jan.-Feb., 2527.
Vacher, H.L. (1997), Introduction: Varieties of carbonate islands and a historical perspective. In
Geology and Hydrogeology of Carbonate Islands. Developments in Sedimentology 54,
eds. H.L. Vacher and T. Quinn, 1-33.
Volker RE, Mariño MA, Rolston DE (1985) Transition zone width in ground water on ocean
atolls. J Hydraul Eng 111(4):659–676
73
Voss, C. I. (1984). A finite-element simulation model for saturated-unsaturated, fluid-densitydependent ground-water flow with energy transport or chemically-reactive single-species
solute transport (No. 84-4369). US Geological Survey,.
Voss, C.I. and A.M. Provost (2010), SUTRA, A model for saturated-unsaturated variable-density
ground-water flow with solute or energy transport. USGS Water-Resources
Investigations Report 02-4231. Reston, Virginia: USGS.
White, I., Falkland, T., Metutera, T., Metai, E., Overmars, M., Perez, P., Dray, A., 2007.
Climatic and human influences on groundwater in low atolls. Vadose Zone J. 6, 581–
590.
WHO (World Health Organization) (1972), International standards for drinking-water. WHO
(U.N.), Geneva.
WHO(World Health Organization), 2009. Guidelines and Manual for Rain Water Harvesting in
Maldives. Male, Maldives: Ministry of Housing Transport and Environment Government
of the Republic of Maldives Technical Support.
Wilkinson C, O. Linden, H. Cesar, G. Hodgson, J. Rubens, and A.E. Strong, 1999. Ecological
and socioeconomic impacts of 1998 coral mortality in the Indian Ocean: An ENSO
impact and a warning of future change? Ambio 28, 188-196.
Winston, Richard B. ModelMuse: a graphical user interface for MODFLOW-2005 and PHAST.
US Geological Survey, 2009.
Woodroffe C. D and Falkland A.C. (1997). Geology and hydrogeology of the Cocos (Keeling)
Islands, Chapter 31, Geology and Hydrogeology of Carbonate Islands, Developments in
Sedimentology 54 (editors Vacher, H.L. and Quinn, T.M., Elsevier, Amsterdam.
Woodroffe, C.D. 2008. Reef-island topography and the vulnerability of atolls to sea-level rise.
Global and Planetary Change 62, 77-96.
Woodworth, P.L. 2005. Have there been large recent sea level changes in the Maldive Islands
Global and Planetary Change 49, 1-18.
Woodworth, P.L., C. Le Provost, L.J. Richards, G.T. Mitchum and M. Merrifield (2002), A
review of sea level research from tide gauges during the World Ocean Current
Experiment. Oceanogr. Mar. Biol., 40 , 1-35.
74
APPENDIX I
Table A 1. Model performance results for monthly rainfall rates in Region 1 and RCP Scenario 4.5, ranking best to
worst according to the total score.
GCM
IPSL-CM5A-MR
MIROC5
GFDL-ESM2M
FIO-ESM
HadGEM2-ES
CSIRO-Mk3-6-0
NorESM1-M
GISS-E2-R p3
IPSL-CM5A-LR
CESM1-CAM5
GISS-E2-H p3
GFDL-CM3
GISS-E2-H p1
CCSM4
GISS-E2-H p2
GFDL-ESM2G
NorESM1-ME
GISS-E2-R p2
GISS-E2-R p1
MIROC-ESM
MRI-CGCM3
MIROC-ESM-CHEM
HadGEM2-AO
bcc-csm1-1
Mean
RE(mm)
-0.04
-0.05
-0.11
0.07
-0.20
-0.16
-0.04
-0.11
-0.16
0.28
-0.10
0.13
-0.09
0.24
-0.06
-0.25
0.00
-0.12
-0.13
0.55
-0.20
0.59
-0.22
0.24
Std
RE
-0.14
0.05
-0.04
0.01
0.02
-0.11
0.17
0.06
-0.43
0.01
0.18
0.18
0.28
0.08
0.29
-0.22
0.16
0.21
0.19
0.46
0.28
0.42
0.15
0.80
NRMSE
Corr
BS
0.87
0.83
0.94
1.09
1.07
0.87
1.15
1.03
0.85
1.04
1.10
1.04
1.18
1.07
1.16
1.02
1.29
1.13
1.15
1.39
1.23
1.41
1.22
1.79
0.57
0.67
0.56
0.41
0.47
0.61
0.45
0.51
0.57
0.55
0.51
0.57
0.49
0.52
0.51
0.45
0.29
0.50
0.47
0.63
0.47
0.62
0.40
0.32
4.19
5.23
2.93
3.79
5.79
5.69
4.83
6.03
5.09
4.61
6.49
4.87
5.81
4.83
6.51
4.23
4.11
7.71
7.65
4.11
14.51
4.81
8.65
12.03
75
S
score
-0.04
-0.05
-0.11
0.07
-0.20
-0.16
-0.04
-0.11
-0.16
0.28
-0.10
0.13
-0.09
0.24
-0.06
-0.25
0.00
-0.12
-0.13
0.55
-0.20
0.59
-0.22
0.24
Kendal Slope
(mm/year)
-0.14
0.05
-0.04
0.01
0.02
-0.11
0.17
0.06
-0.43
0.01
0.18
0.18
0.28
0.08
0.29
-0.22
0.16
0.21
0.19
0.46
0.28
0.42
0.15
0.80
Total
Score
33
34.5
44
50
58
62
65.5
67
68
70
70.5
71.5
74.5
77
77.5
80.5
82
87
88.5
93
95
95.5
111
135
Table A2. Model performance results for monthly rainfall rates in Region 1 and RCP Scenario 6.0, ranking best to
worst according to the total score.
GCM
MIROC5
IPSL-CM5A-MR
GFDL-CM3
CESM1-CAM5
CSIRO-Mk3-6-0
GISS-E2-H p3
FIO-ESM
NorESM1-M
GFDL-ESM2G
HadGEM2-ES
NorESM1-ME
GISS-E2-R p1
GISS-E2-R p2
CCSM4
GISS-E2-H p1
GISS-E2-R p3
IPSL-CM5A-LR
GFDL-ESM2M
GISS-E2-H p2
MIROC-ESM-CHEM
MRI-CGCM3
MIROC-ESM
HadGEM2-AO
bcc-csm1-1
Mean
RE(mm)
-0.09
0.00
0.07
0.24
-0.12
-0.07
0.07
-0.01
-0.19
-0.16
-0.01
-0.08
-0.09
0.24
-0.04
-0.10
-0.13
0.07
-0.11
0.58
-0.22
0.55
-0.23
0.21
Std
RE
0.04
-0.17
0.15
0.01
-0.06
0.25
0.01
0.22
-0.16
0.06
0.16
0.33
0.30
0.11
0.47
0.21
-0.23
0.26
0.45
0.44
0.27
0.41
0.16
0.74
NRMSE
Corr
BS
0.86
0.92
1.01
1.01
0.86
1.16
1.11
1.21
1.01
1.10
1.29
1.24
1.22
1.08
1.34
1.20
0.99
1.40
1.33
1.38
1.25
1.36
1.26
1.67
0.65
0.50
0.56
0.56
0.62
0.49
0.39
0.42
0.45
0.46
0.29
1.24
0.46
0.53
0.46
0.43
0.42
0.25
0.47
0.65
0.46
0.63
0.37
0.38
4.79
4.15
3.45
4.75
6.15
3.81
4.05
5.16
3.29
5.11
3.89
5.91
6.55
5.05
6.09
4.29
4.21
4.01
5.75
5.21
20.95
4.41
10.77
12.57
76
S
score
99
95
111
90
84
103
96
99
103
99
101
94
90
84
99
98
97
107
94
90
80
92
84
84
Kendal Slope
(mm/year)
0.0006
0.0013
0.0026
0.0005
0.0026
0.0027
0.0028
0.0024
0.0046
0.0033
0.0032
-0.0009
0.0009
0.0045
0.0021
-0.0019
0.0089
0.0052
-0.0007
0.0025
0.0019
0.0056
0.0089
0.0055
Total
Score
29
35
36.5
54
55
58
59
64.5
68
69
72.5
74
76
77.5
77.5
79.5
83.5
91.5
93.5
94.5
96
103
113
129.5
Table A3. Model performance results for monthly rainfall rates in Region 1 and RCP Scenario 8.5, ranking best to
worst according to the total score.
GCM
IPSL-CM5A-MR
GFDL-ESM2M
MIROC5
FIO-ESM
CSIRO-Mk3-6-0
CESM1-CAM5
GFDL-CM3
GFDL-ESM2G
GISS-E2-R p3
NorESM1-M
NorESM1-ME
GISS-E2-R p2
GISS-E2-H p3
HadGEM2-ES
CCSM4
IPSL-CM5A-LR
GISS-E2-R p1
GISS-E2-H p1
MIROC-ESM
GISS-E2-H p2
MIROC-ESM-CHEM
HadGEM2-AO
MRI-CGCM3
bcc-csm1-1
Mean
RE(mm)
-0.03
-0.08
-0.04
0.04
-0.13
0.26
0.11
-0.21
-0.10
0.03
0.03
-0.07
-0.05
-0.19
0.24
-0.14
-0.07
-0.04
0.53
-0.09
0.55
-0.28
-0.19
0.12
Std
RE
-0.20
-0.02
0.06
0.01
-0.08
0.04
0.16
-0.23
0.21
0.26
0.26
0.33
0.30
0.02
0.10
-0.42
0.34
0.44
0.39
0.40
0.42
0.07
0.32
0.66
NRMSE
Corr
BS
0.89
0.98
0.84
1.07
0.87
1.01
1.01
0.98
1.21
1.18
1.18
1.23
1.21
1.08
1.09
0.86
1.33
1.31
1.32
1.31
1.35
1.23
1.30
1.66
0.53
0.52
0.67
0.42
0.61
0.58
0.58
0.47
0.42
0.47
0.47
0.48
0.47
0.46
0.52
0.54
0.38
0.47
0.64
0.45
0.65
0.37
0.42
0.31
3.73
3.37
5.01
3.73
5.63
4.53
4.83
3.49
3.99
4.23
4.23
7.11
3.97
6.09
4.99
5.17
5.39
6.15
4.47
6.17
4.73
8.13
16.25
12.65
77
S
score
101
105
97
97
85
92
105
104
96
99
99
95
108
89
84
90
95
97
97
92
82
84
91
86
Kendal Slope
(mm/year)
0.0007
0.0016
0.0059
0.0004
0.0016
0.0021
0.0056
0.0032
0.0012
0.0104
0.0104
0.0015
0.0074
-0.0007
0.0055
0.0064
0.0016
0.0035
0.0039
0.002
0.0001
0.0023
0.0038
-0.0035
Total
Score
30
32.5
42
47.5
52
58.5
61
65
69
69.5
69.5
72
77
77.5
82.5
83.5
88.5
90.5
90.5
93.5
98.5
101
107.5
126.5
Table A4. Model performance results for monthly rainfall rates in Region 2 and RCP Scenario 2.6, ranking best to
worst according to the total score.
GCM
CESM1-CAM5
MIROC5
IPSL-CM5A-LR
CCSM4
GFDL-ESM2G
GISS-E2-R p1
NorESM1-ME
IPSL-CM5A-MR
GISS-E2-R p2
HadGEM2-ES
NorESM1-M
GFDL-CM3
GISS-E2-R p3
FIO-ESM
GISS-E2-H p1
MIROC-ESM
MIROC-ESM-CHEM
MRI-CGCM3
GISS-E2-H p2
GFDL-ESM2M
CSIRO-Mk3-6-0
bcc-csm1-1
GISS-E2-H p3
HadGEM2-AO
Mean
RE(mm)
0.09
-0.12
0.19
-0.14
0.01
0.07
-0.12
0.25
0.08
0.25
-0.06
0.14
0.12
0.03
0.16
0.43
0.43
0.23
0.27
0.26
0.59
-0.11
0.31
0.85
Std
RE
0.26
-0.22
-0.21
0.02
0.17
0.42
0.25
0.16
0.41
0.52
0.45
0.57
0.54
0.30
0.55
-0.04
-0.14
0.90
0.49
0.86
0.49
0.69
0.68
0.85
NRMSE
Corr
BS
1.33
0.97
1.04
1.08
1.34
1.46
1.41
1.37
1.44
1.39
1.57
1.45
1.61
1.49
1.68
1.49
1.46
1.71
1.65
1.78
1.77
1.83
1.86
1.79
0.33
0.47
0.43
0.46
0.24
0.31
0.24
0.29
0.33
0.53
0.22
0.45
0.27
0.19
0.22
0.20
0.17
0.49
0.25
0.42
0.46
0.15
0.21
0.42
3.69
4.33
3.93
3.95
4.07
3.97
3.63
4.35
4.49
6.03
4.75
4.93
4.93
3.83
3.35
4.63
4.79
4.99
4.57
4.93
5.81
11.15
4.01
4.99
78
S
score
111
104
102
104
109
105
108
110
96
88
99
95
95
101
108
93
85
90
96
113
75
100
99
91
Kendal Slope
(mm/year)
-0.002
0.0023
0.0007
0.0046
0.0022
-0.0033
0.0034
0.0054
0.004
-0.0002
0.0028
0.0029
-0.0011
0.0062
0.0018
-0.0025
-0.001
0.0031
0.0056
0.0105
0.0051
-0.0083
0.0121
0.0109
Total
Score
34.5
39
42.5
43
43.5
53
57.5
64
64.5
67.5
74
75
75
75.5
79
79.5
81.5
93.5
98
102
103
113
118.5
119
Table A5. Model performance results for monthly rainfall rates in Region 2 and RCP Scenario 4.5, ranking best to
worst according to the total score.
GCM
CESM1-CAM5
GISS-E2-R p1
IPSL-CM5A-LR
GISS-E2-R p2
GFDL-ESM2G
CCSM4
MIROC5
GISS-E2-H p1
GISS-E2-R p3
FIO-ESM
IPSL-CM5A-MR
MRI-CGCM3
GISS-E2-H p2
NorESM1-M
GFDL-CM3
HadGEM2-ES
GISS-E2-H p3
MIROC-ESM
NorESM1-ME
MIROC-ESM-CHEM
CSIRO-Mk3-6-0
GFDL-ESM2M
HadGEM2-AO
bcc-csm1-1
Mean
RE(mm)
0.00
0.05
0.19
0.07
0.01
-0.13
-0.11
0.15
0.11
0.02
0.24
0.20
0.20
-0.08
0.17
0.26
0.20
0.44
-0.10
0.46
0.57
0.29
0.28
-0.09
Std
RE
-0.11
0.06
-0.18
0.12
0.10
0.03
-0.19
0.25
0.18
0.22
0.22
0.91
0.13
0.36
0.61
0.56
0.27
-0.02
0.36
-0.15
0.47
1.09
0.89
0.69
NRMSE
Corr
BS
0.92
1.10
1.03
1.17
1.26
1.08
0.96
1.29
1.28
1.43
1.40
1.70
1.26
1.50
1.48
1.44
1.38
1.53
1.49
1.47
1.74
2.05
1.76
1.82
0.52
0.43
0.45
0.40
0.28
0.46
0.47
0.39
0.34
0.17
0.30
0.50
0.37
0.22
0.47
0.52
0.33
0.16
0.24
0.20
0.45
0.35
0.46
0.16
3.81
100
3.99
4.39
4.05
3.63
4.07
3.89
4.01
3.47
2.79
5.27
4.39
3.92
4.37
5.77
4.43
5.55
4.41
5.33
6.35
3.81
6.33
9.43
79
S
score
103
100
103
99
106
106
104
97
101
107
119
92
89
102
111
84
89
81
106
81
75
118
79
102
Kendal Slope
(mm/year)
0.0028
-0.0006
-0.0002
0.0008
0.0022
0.0064
0.0043
0.0005
0.0034
0.0047
0.0033
-0.0012
0.0057
0.0031
0.005
0.001
0.0035
0.0002
0.0057
0.001
0.0038
0.0173
0.0049
-0.0086
Total
Score
24.5
43
46.5
48
48.5
50
51.5
64
68
70
73
79
79.5
80
80.5
81
89
90
91
93.5
109
110
111
113
Table A 6. Model performance results for monthly rainfall rates in Region 2 and RCP Scenario 6.0, ranking best to
worst according to the total score.
GCM
CESM1-CAM5
GFDL-ESM2G
MIROC5
CCSM4
GFDL-CM3
NorESM1-ME
FIO-ESM
MIROC-ESM-CHEM
GISS-E2-H p2
GISS-E2-R p3
IPSL-CM5A-LR
GISS-E2-R p2
GISS-E2-H p3
NorESM1-M
GISS-E2-R p1
MIROC-ESM
MRI-CGCM3
HadGEM2-ES
GFDL-ESM2M
HadGEM2-AO
IPSL-CM5A-MR
bcc-csm1-1
GISS-E2-H p1
CSIRO-Mk3-6-0
Mean
RE(mm)
0.01
-0.04
-0.12
-0.13
0.13
-0.11
0.03
0.44
0.17
0.12
0.21
0.07
0.21
-0.07
0.13
0.46
0.18
0.28
0.25
0.19
0.49
0.06
0.20
0.61
Std
RE
-0.12
0.06
-0.18
0.04
0.51
0.33
0.30
-0.08
0.41
0.45
0.17
0.41
0.63
0.40
0.50
0.02
0.91
0.62
0.78
0.83
0.22
1.00
0.56
0.53
NRMSE
Corr
BS
0.93
1.02
1.02
1.09
1.42
1.48
1.52
1.46
1.49
1.54
1.34
1.49
1.70
1.53
1.50
1.53
1.74
1.51
1.71
1.74
1.81
2.04
1.76
1.82
0.51
0.52
0.42
0.45
0.44
0.22
0.14
0.24
0.30
0.26
0.31
0.28
0.29
0.22
0.35
0.23
0.46
0.50
0.41
0.39
0.05
0.21
0.15
0.46
3.53
3.03
3.49
3.69
4.33
4.17
3.01
4.51
3.53
4.59
4.41
5.33
3.31
4.98
3.67
4.91
5.91
5.05
3.51
6.47
6.65
8.49
3.37
7.17
80
S
score
103
109
107
106
102
106
110
89
103
95
98
89
109
94
101
90
93
90
120
82
86
102
104
71
Kendal Slope
(mm/year)
0.0033
0.0074
0.0018
0.0053
0.0013
0.0025
0.0057
-0.0011
-0.0053
-0.0015
0.0061
0.0036
0.001
0.0023
0.0066
0.0026
0.0001
0.0068
0.0086
-0.0023
-0.0007
0.0057
0.0089
0.0087
Total
Score
30
33.5
38
46
57.5
62.5
64.5
67
68
72
72.5
75.5
77
77.5
79
83
84.5
92
92.5
93
102
106.5
107
114.5
Table A7. Model performance results for monthly rainfall rates in Region 2 and RCP Scenario 8.5, ranking best to
worst according to the total score.
GCM
CESM1-CAM5
MIROC5
CCSM4
IPSL-CM5A-MR
GISS-E2-R p1
GFDL-ESM2G
FIO-ESM
GFDL-CM3
IPSL-CM5A-LR
NorESM1-ME
GISS-E2-H p3
MRI-CGCM3
GISS-E2-H p2
MIROC-ESM-CHEM
GFDL-ESM2M
MIROC-ESM
GISS-E2-R p3
NorESM1-M
GISS-E2-H p1
HadGEM2-ES
GISS-E2-R p2
HadGEM2-AO
CSIRO-Mk3-6-0
bcc-csm1-1
Mean
RE(mm)
0.03
-0.11
-0.13
0.23
0.10
0.02
0.01
0.13
0.22
-0.07
0.21
0.18
0.18
0.43
0.19
0.46
0.16
-0.03
0.14
0.26
0.11
0.21
0.59
Std
RE
-0.10
-0.23
0.01
0.03
0.44
0.09
0.21
0.51
-0.11
0.37
0.48
0.82
0.54
-0.14
0.77
-0.03
0.46
0.44
0.51
0.65
0.41
0.82
0.41
-0.09
0.47
NRMSE
Corr
BS
0.98
0.95
1.10
1.32
1.49
1.30
1.43
1.42
1.15
1.49
1.58
1.61
1.65
1.43
1.74
1.49
1.58
1.54
1.64
1.50
1.55
1.71
1.73
0.47
0.48
0.43
0.25
0.31
0.23
0.17
0.44
0.37
0.24
0.29
0.50
0.25
0.23
0.34
0.26
0.24
0.24
0.21
0.52
0.22
0.42
0.46
0.13
3.77
4.09
3.89
3.73
2.69
4.29
3.65
4.51
4.17
3.55
3.41
4.53
3.87
4.57
3.29
5.65
4.13
4.33
3.61
5.35
4.25
5.71
6.47
S
score
101
104
105
107
116
105
107
100
99
110
102
98
105
92
123
81
96
96
99
88
99
87
73
7.87
81
1.89
81
Kendal Slope
(mm/year)
0.0064
0.0023
0.0062
-0.0007
0.0015
0.0041
0.0041
0.0016
0.0045
0.0081
0.0002
0.0001
-0.002
-0.0028
0.0022
0.0024
0.0013
0.0075
0.0009
0.0038
0.0084
0.0008
0.0047
Total
Score
43.5
44
48
49.5
53.5
56.5
61
67.5
68
68.5
72.5
77
78.5
80
83.5
83.5
84.5
86.5
87
90
91.5
98.5
104.5
0.0104
117
Table A8. Model performance results for monthly rainfall rates in Region 3 and RCP Scenario 2.6, ranking best to
worst according to the total score.
GCM
CESM1-CAM5
NorESM1-ME
CCSM4
MIROC-ESM
IPSL-CM5A-LR
MIROC5
MIROC-ESM-CHEM
FIO-ESM
NorESM1-M
GISS-E2-R p1
IPSL-CM5A-MR
GFDL-CM3
HadGEM2-ES
HadGEM2-AO
GISS-E2-H p1
MRI-CGCM3
GFDL-ESM2G
GFDL-ESM2M
GISS-E2-R p3
GISS-E2-H p2
CSIRO-Mk3-6-0
GISS-E2-R p2
GISS-E2-H p3
bcc-csm1-1
Mean
RE(mm)
0.26
0.01
0.24
0.54
-0.16
-0.07
0.58
0.07
-0.03
-0.09
-0.01
0.09
-0.19
-0.23
-0.09
-0.21
-0.22
-0.03
-0.09
-0.06
-0.12
-0.11
-0.05
0.25
Std
RE
-0.24
0.22
-0.34
-0.23
-0.14
-0.33
-0.23
-0.18
0.27
0.56
0.17
0.57
1.71
0.75
0.84
0.92
0.48
0.75
0.64
0.72
0.55
0.57
0.88
1.06
NRMSE
Corr
BS
1.07
1.53
1.04
1.24
1.30
1.13
1.21
1.33
1.70
1.83
1.57
1.68
1.71
1.96
2.09
2.02
1.81
2.00
2.02
2.07
2.08
1.89
2.25
2.19
0.28
0.14
0.27
0.04
0.15
0.13
0.08
0.01
0.00
0.22
0.08
0.29
0.20
0.18
0.22
0.18
0.07
0.22
0.13
0.14
0.18
0.13
0.13
0.14
3.73
2.67
5.39
2.79
4.57
3.49
3.43
4.13
2.83
3.21
3.45
3.53
3.71
3.97
4.22
4.21
3.79
3.59
4.05
5.21
6.63
4.71
3.91
16.09
82
S
score
106
111
100
110
94
107
109
95
106
105
111
111
103
104
100
100
103
103
94
84
73
89
102
68
Kendal Slope
(mm/year)
0.0051
0.0035
0.0028
-0.0007
0.0015
-0.0015
-0.0019
0.0031
0.0033
0.0046
-0.004
-0.0036
-0.0009
0.0073
0.0036
-0.0011
-0.0063
-0.0071
0.0024
0.0045
0.0002
0.0137
0.0011
-0.0078
Total
Score
35
38
40.5
49
56
56.5
57
60
61
62.5
66.5
72.5
79.5
82
86.5
88
91.5
92
97
97.5
101
103.5
108.5
109
Table A 9. Model performance results for monthly rainfall rates in Region 3 and RCP Scenario 4.5, ranking best to
worst according to the total score.
GCM
CESM1-CAM5
CCSM4
MIROC-ESM
MIROC-ESM-CHEM
MIROC5
IPSL-CM5A-LR
IPSL-CM5A-MR
NorESM1-ME
HadGEM2-ES
NorESM1-M
GISS-E2-R p2
GISS-E2-H p2
GISS-E2-R p1
FIO-ESM
GFDL-ESM2G
GISS-E2-H p1
GISS-E2-R p3
MRI-CGCM3
GISS-E2-H p3
GFDL-CM3
bcc-csm1-1
HadGEM2-AO
GFDL-ESM2M
CSIRO-Mk3-6-0
Mean
RE(mm)
0.00
-0.06
0.01
0.02
0.06
0.18
0.23
0.11
0.22
0.17
0.28
0.31
0.32
0.11
0.28
0.33
0.36
0.14
0.35
0.22
-0.07
0.21
0.36
0.51
Std
RE
-0.23
-0.33
-0.20
-0.20
-0.27
-0.16
0.25
0.25
0.57
0.21
0.38
0.54
0.30
-0.24
0.51
0.53
0.29
0.97
0.53
0.61
1.00
0.74
0.70
0.45
NRMSE
Corr
BS
1.05
1.05
1.25
1.18
1.15
1.26
1.63
1.53
1.74
1.59
1.68
1.86
1.72
1.27
1.89
1.81
1.74
2.08
1.86
1.81
2.08
1.93
1.96
2.04
0.32
0.27
0.04
0.14
0.15
0.19
0.09
0.12
0.22
0.04
0.21
0.17
0.15
0.03
0.06
0.25
0.16
0.17
0.21
0.19
0.17
0.16
0.24
0.20
4.05
4.45
2.93
3.49
2.67
4.57
3.15
2.69
3.73
3.11
3.79
3.39
3.95
4.33
3.37
4.69
4.25
4.41
4.57
5.37
10.51
4.25
5.11
6.03
83
S
score
101
102
107
107
114
95
111
111
97
107
97
98
99
100
101
89
93
96
92
97
84
98
91
76
Kendal Slope
(mm/year)
0.0018
0.0019
0.0024
0.0019
-0.0001
0.0012
0.004
-0.0082
0.002
-0.0029
0.0056
0.0034
0.002
-0.0006
0.0032
0.0035
0.0036
0.003
0.0012
0.0006
0.0012
-0.0044
0.0063
0.0057
Total
Score
30
42
42.5
44
55
58
61.5
68
69.5
71
75
76.5
77
78.5
79
79
81
88.5
93
96
101.5
104.5
108
108.5
Table A10. Model performance results for monthly rainfall rates in Region 3 and RCP Scenario 6.0, ranking best to
worst according to the total score.
GCM
CESM1-CAM5
IPSL-CM5A-LR
CCSM4
MIROC-ESM
MIROC5
NorESM1-M
MIROC-ESM-CHEM
NorESM1-ME
FIO-ESM
GFDL-CM3
HadGEM2-ES
GISS-E2-R p1
GISS-E2-H p2
GFDL-ESM2G
CSIRO-Mk3-6-0
GISS-E2-R p3
GISS-E2-H p1
HadGEM2-AO
MRI-CGCM3
GISS-E2-H p3
GFDL-ESM2M
GISS-E2-R p2
bcc-csm1-1
IPSL-CM5A-MR
Mean
RE(mm)
0.01
0.18
-0.06
0.02
0.06
0.20
-0.02
0.11
0.12
0.22
0.17
0.31
0.33
0.26
0.52
0.34
0.37
0.16
0.12
0.41
0.32
0.28
0.03
0.07
Std
RE
-0.21
0.28
-0.30
-0.23
-0.27
0.27
-0.17
0.19
-0.20
0.54
0.58
0.62
0.73
0.51
0.55
0.59
0.86
0.82
1.16
0.91
0.65
0.63
1.26
1.41
NRMSE
Corr
BS
1.10
1.53
1.06
1.22
1.17
1.62
1.28
1.51
1.29
1.69
1.79
1.93
1.98
1.87
2.13
1.94
2.18
1.97
2.28
2.26
1.99
1.97
2.34
2.46
0.25
0.20
0.28
0.07
0.13
0.08
0.02
0.09
0.03
0.26
0.15
0.15
0.21
0.06
0.18
0.14
0.14
0.16
0.12
0.15
0.12
0.08
0.13
0.16
4.15
2.91
5.09
3.31
3.17
2.97
3.43
2.73
3.47
3.93
2.59
3.33
4.57
4.01
6.95
4.59
3.95
4.77
4.77
4.65
4.33
3.73
9.41
10.75
84
S
score
100
112
98
108
109
113
109
116
105
107
109
110
97
107
68
95
98
95
100
91
100
98
84
83
Kendal Slope
(mm/year)
0.003
0.0034
0.0017
0.0019
-0.0005
0.0027
-0.0032
-0.0058
0.001
-0.0012
-0.0024
0.0063
0.0032
-0.001
0.0045
0.0001
0.0049
-0.0091
-0.0012
0.0055
-0.004
0.0127
0.0109
0.0117
Total
Score
26
38
43
45
48
55
57.5
58.5
61
63
67
72.5
77
85
89
91
95.5
96
100.5
102.5
103.5
103.5
106.5
107.5
Table A11. Model performance results for monthly rainfall rates in Region 3 and RCP Scenario 8.5, ranking best to
worst according to the total score.
GCM
CESM1-CAM5
MIROC-ESM
CCSM4
MIROC-ESM-CHEM
NorESM1-ME
HadGEM2-ES
GFDL-CM3
IPSL-CM5A-LR
FIO-ESM
IPSL-CM5A-MR
MIROC5
HadGEM2-AO
GISS-E2-R p1
NorESM1-M
MRI-CGCM3
CSIRO-Mk3-6-0
GFDL-ESM2G
GISS-E2-H p1
GISS-E2-R p2
bcc-csm1-1
GISS-E2-H p2
GISS-E2-R p3
GFDL-ESM2M
GISS-E2-H p3
Mean
RE(mm)
0.00
0.03
-0.06
0.01
0.17
0.21
0.24
0.18
0.11
0.21
0.04
0.24
0.25
0.18
0.15
0.51
0.24
0.32
0.25
-0.08
0.34
0.34
0.32
0.37
Std
RE
-0.25
-0.21
-0.33
-0.19
0.20
0.66
0.53
-0.14
-0.17
0.15
-0.32
0.95
0.55
0.16
1.00
0.51
0.42
0.80
0.54
1.08
0.81
0.63
0.78
0.91
NRMSE
Corr
BS
1.08
1.24
1.04
1.21
1.54
1.75
1.70
1.29
1.32
1.55
1.12
2.02
1.79
1.57
2.16
2.09
1.78
2.05
1.88
2.16
2.09
2.00
2.10
2.24
0.25
0.05
0.29
0.12
0.10
0.29
0.26
0.16
0.00
0.08
0.16
0.28
0.19
0.03
0.11
0.18
0.06
0.19
0.07
0.16
0.17
0.11
0.10
0.13
4.01
2.65
5.11
3.25
2.95
3.45
4.13
4.41
3.35
2.89
4.61
4.49
3.31
3.85
4.39
6.59
3.87
4.57
4.35
10.79
4.85
4.13
4.87
4.37
85
S
score
102
112
95
106
113
101
107
96
107
115
100
98
107
99
98
77
99
86
89
79
87
96
100
98
Kendal Slope
(mm/year)
0.0022
0.0045
0.002
-0.0018
0.0023
0.004
0.0042
0.001
-0.0001
-0.0017
-0.0043
0.0038
-0.0027
-0.0044
0.0039
0.0025
-0.0039
0.0007
0.0073
-0.0007
0.0012
-0.0029
-0.0009
-0.0024
Total
Score
32
47
47.5
48.5
49.5
53
55.5
56.5
60
61.5
64.5
74.5
77.5
81
84.5
90.5
91.5
96
96.5
100.5
101.5
104
106.5
114
Table A12. Average lens thickness under the center of the island for each island width and geographic region,
across all accepted GCMs from the RCP8.5. The first set of values is averages through the years 2011-2030, and the
second set is for the years 2031-2050. The GCM index corresponds to the order listed in Table 6.
Last 20 Years (2031-2050)
First 20 Years (2011-2030)
Average Lens
Thickness (m)
GCM
1
GCM
2
GCM
3
GCM
4
GCM
5
GCM
6
GCM
7
Mean
GCM
1
GCM
2
GCM
3
GCM
4
GCM
5
GCM
6
GCM
7
Mean
Island Width (m)
600
R3
R1
R2
R1
200
R2
R3
R1
400
R2
R3
R1
1100
R2
R3
0.23
0.33
0.81
1.96
2.22
4.00
4.92
5.46
8.81
11.27
11.66
12.92
0.23
0.19
0.93
2.06
2.27
4.02
5.15
5.59
8.70
11.45
11.76
12.99
0.21
0.36
0.83
2.11
2.65
4.31
5.25
6.28
9.29
11.50
12.03
13.13
0.30
0.28
2.21
2.66
5.41
6.28
11.59
12.09
0.19
0.23
2.23
2.86
5.44
6.62
11.61
12.19
0.27
0.42
2.38
2.89
5.73
6.74
11.75
12.22
0.19
0.23
R1
0.45
0.32
R2
0.85
R3
2.19
R1
2.93
2.64
R2
4.11
R3
5.81
5.39
R1
6.81
6.26
R2
8.93
R3
11.76
11.56
R1
12.33
12.04
R2
13.01
R3
0.26
0.35
0.96
2.03
1.82
3.95
5.16
4.62
8.66
11.38
11.14
12.98
0.20
0.20
0.80
2.13
2.59
4.33
5.30
6.16
9.31
11.54
12.00
13.14
0.22
0.34
0.95
2.22
2.69
4.39
5.41
6.30
9.37
11.61
12.10
13.16
0.54
0.30
2.28
2.72
5.52
6.37
11.68
12.12
0.24
0.19
2.67
3.13
6.31
7.17
11.97
12.44
0.38
0.67
2.84
3.34
6.59
7.60
12.08
12.48
0.43
0.59
3.20
3.67
7.18
8.13
12.30
12.64
0.32
0.38
2.48
2.85
5.93
6.62
11.79
12.13
2.93
0.90
4.22
86
9.11
13.09
APPENDIX II
Figure B1. Comparison of historical rainfall data time series plots from worst GCMs for RCP.2.6
87
Figure B2. Time series plot of lens thickness for each accepted GCM in Region 2
Figure B3. Time series plot of lens thickness for each accepted GCM in Region 3
88
Figure B4. Boxplots of best five GCMs for each scenario in each region for 200m island
89
Figure B5. Comparison of scenario 26 boxplot of fresh lens for different three regions for 200 m island
90
Figure B6. PDF plot of fresh lens between GCMs in 200 m island
91
Figure B7. Boxplots of best five GCMs for each scenario in each region for 400m island
92
Figure B8. Comparison of scenario 26 boxplot of fresh lens for different three regions for 400 m island
93
Figure B9. PDF plot of fresh lens between GCMs in 400 m island
94
Figure B10. Boxplots of best five GCMs for each scenario in each region for 600m island
95
Figure B11. Comparison of scenario 26 boxplot of fresh lens for different three regions for 600 m island
96
Figure B12. PDF plot of fresh lens between GCMs in 600 m island
97
Figure B13. Comparison of scenario 26 boxplot of fresh lens for different three regions for 1100 m island
98
Figure B14. Boxplots of best five GCMs for each scenario in each region for 1100m island
99
Figure B15. PDF plot of fresh lens between GCMs in 1100 m island
100
Figure B16. Lens thickness PDF plot for 200m islands
101
Figure B17. Lens thickness plot for 600 m islands
102